The last thing this needs is another interpretation of the word "purpose" but I'm doing it anyway. Though I agree that POSIWID is useless language, it's always been my understanding that "purpose" meant like... intent-of-the-thing-itself. Grammatically, the same usage as "it is my purpose [in writing you] to get this job." Very behaviorism-analogous in that it doesn't suppose the system has to have any kind of subjectivity to have incentives (or at least attractors).
Like, it obviously isn't the _design intention_ of the hospital to cure two-thirds of cancer patients. But the hospital-as-system will very "happily" lope along in its established niche, which is in curing exactly the number of people that can be cured without any expenditure TOO too great, as has been determined by the surrounding pressures to date. It will seem to resist you when you try to perturb it out of that niche. Hopefully there's a concise way to say that without being violently reductive or just misleading about the point.
It has incentives to improve that rate and also has incentives to keep costs manageable. It is the tradeoff between cures and costs that is the purpose, on this account, not the precise number of cures.
I am skeptical, however, that if someone found a (cost-free) way to improve outcomes by 1% that somebody else would immediately propose saving whatever money that allowed and returning to 66%, which feels like the implication of "loping along happily at 66%."
That wasn't the implication. It could be someone's implication, and there are situations where it could actually happen I guess, but it's not what I said or meant. I actually said "as has been determined by the surrounding pressures to date" which would seem to account for the possibility of the cost curves changing.
Why would an increased cure rate necessarily save money?
For my part, I'd say that a given hospital may not be strongly motivated to improve to 67% cure rate if 66% is the standard of care. There may be exceptions to this, in the same way that you can buy aggregated food ingredients or more expensive but higher quality single-source ingredients. But most people buy the aggregated stuff.
However, if the standard of care was 80% cure rate, then there might be stronger incentives for a given hospital to improve. Many doctors are happy blaming a patient's symptoms on obesity or hypochondria or whatever if they're able to get away with it. (I'm not talking from personal experience, but from the experiences of those around me.)
Saying the tradeoff between cures and costs is the purpose of the hospital is like saying that the tradeoff between gas mileage and engine power is the purpose of a car.
Is that supposed to be a problem? I suppose it's clear that there's at least a few other parts of the purpose of a car - it's a place to store your goods while you're going around town, it aims to carry a certain number of people, it projects a lifestyle, it feels comfortable. But is it wrong to say that balancing these various interests is the purpose of a car?
Sure, the system has incentives to do that, in that the people who comprise it want to accomplish that. And it also has costs/norms/limitations that regulate those incentives. Maybe nobody internal to the system is actually happy about the result, but the system itself is almost tautologically in... for lack of a better phrase, its locally most-selectable mode of existence.
There exists a system that incentivizes 66% (if that is in fact a steady state operating point). That system is not "the hospital." You need to grow the system to include payment methods and scales, and probably the incentives that determine how insured people are (and undoubtedly many more). Obviously, if there is an equilibrium, then there is a system for which that is the equilibrium.
You need a reason to assign the "desire for equilibrium" as a characteristic to any arbitrary slice of the pie. Just saying that the hospital wants to be at 66% screams, "I am prejudiced against hospitals and looking for an excuse to vent." Anyone inflicted with the same flavor of "woke" may nod their heads in agreement but everyone else will be puzzled, at best.
I don't know who you're replying to that has such strong "woke" (lol) motivations to assign blame, but it isn't me. I think anyone who's studied dynamical systems seriously will accept hospitals as an illustrative example of a large system that's hard to perturb out of a stable mode. You can pick a different one if you like, but I don't see us getting anywhere.
My use of "woke" was more to weaken "woke" than to assign blame. Nonetheless, perhaps I am overreacting to your original message. To me, I essentially agree with (what I take to be) Scott's original position that POSIWID seems to be an excuse for posturing. That this often is often directed at a subset of "woke" readers who are essentially cognoscenti on the particular issue. Apologies if I misread your intent.
The hospital delivers a 'standard of care.' If there's a terminal patient, the hospital will very likely *not* advocate for some radical treatment that has a small chance of improving their outcome. There are going to be incentives other than just for 'improving care.' More strongly, I think it's fair to argue that a hospital may very well *not* have particularly powerful incentives to do better than the average hospital.
A grocery store may very well sell less tasty produce if it means that the unripe produce is less likely to go bad in transit or after arrival, so long as the less tasty produce still sells.
I mean, I agree with Scott's argument about the misuse of POSIWID. But I also agree with a fairly cynical take towards institutions that holds that the incentives incumbent upon a system will be more diverse and self serving than just its mission statement.
Except it needs MY interpretation: Purpose is never singular. There is always more than one of them in action. And they frequently (usually?) aren't aligned.
ISTM that acknowledging this would resolve many of the disagreements.
This has been my problem with this entire discussion.
'Purpose' as a noun has two common definitions, and both involve human agency.
One is as 'intention', as you note. Certainly no system or other inanimate or abstract noun can have a 'purpose' in this sense. The hospital doesn't *intend* to cure 66% of patients, or 99%, or any thing else. The hospital doesn't have any purpose in this sense. It is not purposeful. Nor is the financial system, or a factory, or any other human made system. Not even AI, at this point.
The other common definition is 'purpose' as the reason something is done or created. Note that this also requires human agency, and reflects the intention of the humans who do or create the thing. A healthy beehive in the wild is an excellent system, but we would never talk about the 'purpose' of the beehive as being to provide a home for bees and place for them to make honey. That is how it functions, but it has no purpose because there is no human agency involved. We can say however that the purpose of an artificial behive is to create a similar system in a context in which the honey can be collected.
No system can have a 'purpose' in the sense of the first definition. Any human made system can have a purpose in the second sense. The 'purpose' of the hospital is whatever the creators or maintainers 'intend' for it to do - care for patients, make money, employ doctors. Maybe it does some of these things better than others, but there is no way one can look at the outputs of the system to determine the 'purpose' - one can only look at the human agency behind the endeavor.
When you say, "certainly no system or other inanimate or abstract noun can have a 'purpose' in this sense [intention]," I'm not sure I'm willing to take that for granted... not at 100% of face value, anyway. Certainly it would be anthropomophizing more than I'd like to say, "yes, I do believe those things literally have volition."
(This is somewhat independent of the question whether that was actually the definition employed in POSIWID, which I'm not 100% certain of. But I think it's interesting and important.)
Where's the boundary between dumb feedback and "intention?" Is the latter something we're privileging unnecessarily? Are there things that are essentially "intention" but that humans wouldn't recognize as such?
Also, you say, "we would never talk about the 'purpose' of the beehive as being to provide [...]," and that I don't admit at all. In that framework you can absolutely talk about animal behaviors as having particular functions, and the satisfaction of those functions as "the reason something is done or created." That doesn't imply any kind of cosmic purpose, it's just a straightford etiology.
Whether that's a working framework, whether that "reason" is monolithic, etc. are all valid questions. But people can and do talk about "purposes" of non-human activities in descriptive and consistent ways.
There’s a whole philosophy of biology literature that is relevant here on the concept of function. Roughly speaking, function can be understood as (1) intended function, (2) function as that which was selected for by a quasi-Darwinian process, and (3) function as the role something plays in sustaining a system of which it is a part. Only if we regard intended function as the sole gloss on function do we get the paranoid or ridiculous result your criticisms are directed at. Something can be “performing its function” without performing according to anyone’s intentions.
For (3), it’s relevant that a social system is usually sustained due to lots of intentional actions of people. So someone has to have intentions that partially sustain the behavior of the system. But often lots of people have different intentions, that each sustain parts of the behavior of the system, and it’s useful to think of this complex of intentions as together forming a social function. The function of the bus system is to provide transportation while avoiding impacting the “neighborhood character” of various neighborhoods while not spending too much money and while fitting with someone’s ideological commitments about self-reliance. (If you don’t consider all these conflicting goals, you can’t explain why the bus system keeps expanding and contracting in the particular ways it does.)
> If we both try to predict the behavior of Iranian intelligence, and I’m allowed to use the hypothesis “their purpose is preventing Israeli infiltration”, and you’re not allowed to use that hypothesis, I will consistently outpredict you.
Are you sure you'd outpredict someone who models Iranian intelligence as primarily focused on:
(1) suppressing domestic dissent,
(2) bureaucratic self-preservation,
(3) preemptive counter-elite containment,
(4) serving as a jobs program for regime loyalists,
and a distant (5) concern with preventing Israeli infiltration?
Fake rules: nobody ever agreed to them. If it does all five, in that level of priority, then POSIWID is allowed to use all five. Because the purpose is what it does, and one of the things it does is the stated purpose. Just to a lesser extent.
The distinction is that it's not the main priority. If the system is given an opportunity to sacrifice a non-negligible amount of infiltration defense for another of the priorities then it will do so, because they're higher on the priority list.
If Not-POSIWID is also allowed to use all five AND arrange them in that order of priority and make the same predictions then it's actually just POSIWID in denial.
Did you read the article? The hypothesis is that Iranian intelligence mostly fails at 5, and thus POSIWID would say that its purpose can not be 5.
I think that there are three positions for predictors here.
(A) Strict Non-POSIWID: you are only allowed to look at the mission statement
(B) Sane position: you are allowed at the mission statements, the incentives of individual actors, and the outcomes.
(C) Strict POSIWID: you are only allowed to look at the outcomes
(D) Insane POSIWID: Like (C), but you get to cherry pick one particular aspect of the outcomes ("the justice system exists to let rapists walk free")
Nobody is arguing for the (A). Few people argue for (C) in good faith (see below). (B) allows you to look at the outcomes, the incentives and then wonder if a particular aspect of the outcome might be intended or not. The disagreement is mostly over the semantics of POSIWID, with some people claiming that POSIWID implies (C) and others claiming that Non-POSIWID implies (A). It is People's Front of Judea vs Judean People's Front -- all semantics. (Also, Scott is right on POSIWID meaning (C). Long live the PFJ!)
Regarding (B) vs (C):
Consider alchemists who claim to search for the philosopher's stone, but accidentally discover chemical reactions instead. Per POSIWID, their purpose is the latter (or perhaps some unrelated purpose like inhaling mercury vapors, or finding enlightenment).
I would claim that Non-POSIWID outperforms POSIWID in predictiveness when it comes to how an alchemist would react to synthesizing the stone. POSIWID would be like "nah, that was never the point of alchemy at all. I wonder what happens if I mix phosphorus with vitriol, though."
I don’t care whether anyone endorses (C) in good faith. I just think it’s a productive way to analyze functions of systems like this. But actually look at what it does, not a caricatured incomplete version of it. Using some items from (B) might help you identify outcomes of the system that seem negligible but actually are important in understanding why it is the way it is.
I think it's more accurate to consider this in the form of a continuum. A is 0% POSIWID. C is 100% POSIWID, both are insane extremes. B is obviously true, but only because it logically encompasses all possible worlds from 0.00001% to 99.999999%.
10% POSIWID being the position that organizations almost always are honest and genuinely doing their best, and incentive misalignments sometimes happen but rarely, so you can almost always believe them when they speak. While 90% POSIWID would be the position that people maybe occasionally care about their stated goals, but most failures are incentive-driven, so you can almost always ignore their words and look at actions alone.
Personally, I think the phrase "The Purpose of a System is What it Does" is a slightly ambiguous but useful claim either meaning either that the best point to be is somewhere closer to 100% than to 0%, or that that's true in the particular context where the phrase is being used, or simply that you (or the average person) should be closer than you are right now. Pretty much everyone here lies in B, even the "strict POSIWID" aren't literally at 100%, they're just close enough, probably somewhere in the 90s, that it rounds that way. Everyone here is somewhere in B, just the slightly higher people would like to linguistically claim POSIWID as describing B, or at least the half that they lie in, while the slightly lower people want to consign the phrase POSIWID to the strawman land of C that basically nobody holds. Which is not useful because nobody lives there and then it won't describe anyone's true beliefs.
This actually hits on one of my biggest problems with POSIWID: systems generally do more than one thing, and POSIWID gives us no way to differentiate between them. Not-POSIWID does allow us to differentiate them, designating some as "goals", others as "means", and others still as "side-effects" but, by a strict interpretation of POSIWID, we're not allowed to do that.
This is why I insist that purpose is never singular. But if you collect all the purposes of the system into one category, then you're allowed to collect all the results of the system into one result. (I just don't think that's a useful way of proceeding.)
In Scott's initial example, the assumption is that the system /doesn't/ seem to do #5; hence, POSIWID can't use it.
>If Not-POSIWID is also allowed to use all five AND arrange them in that order of priority and make the same predictions then it's actually just POSIWID in denial.<
Whence the priority list in POSIWID (esp. such that using it is isomorphic to using POSIWID)?
I'd go farther than denying that such a priority list is a unique output of POSIWID, and say that it may not even be an output at all, in a literal reading (& even aside from the problem of #5): you just get a list of all outcomes achieved by the system with no way to differentiate or rank them. Granted, it might be a reasonable extension to add something like "POSIWID—/the most/"...
...but, while "the purpose of a system is what it prioritizes" might indeed be an improvement on the original,¹ I'm not sure this is how it is usually interpreted/used (cf. the examples of POSIWID "in the wild" in the first post).
I.e., Scott seems to be arguing not against "here's how the system uses its resources to trade off between X of outcome 1 vs. Y of outcome 2 vs. ...C of outcome 26; thus, it appears the main goals are, in order: outcomes #[...]", as a useful type of analysis; nor do I read him as advocating for the idea that a system cannot have more than one purpose or do more than one thing; rather, the argument is against a literal reading of "outcome achieved = purpose".
--------------------------------
¹: (I'm also not sure even the improved/priority-based version is so good—I think one can invent situations wherein it too leads to absurdities. E.g., suppose two outcomes are entangled, such as "CO2 production" & "people transported", to borrow the "public transit" illustration: the system appears to prioritize both equally—which, then, is the system's "purpose"? The answer's obvious, of course; but to reach it, it seems to me that we must fall back on to the consideration of intention.)
That just seems plain wrong. That’s a nice exaggeration of the situation, that might be politically useful to claim, and might be helpfully shocking enough to get people to see how far it falls from the ideal statement of 5 as only goal. But the system actually does do lots of little pieces of 5, even if not nearly as effectively as a system designed only for 5 would.
Different parts of the various Iranian intelligence agencies have different purposes.
(I would have recommended a simpler example than this one to avoid complexity issues of multi-purpose organizations.)
So, yes, a huge part of their overall mission is domestic security and keeping public dissent in check. And 2-4 are classics in the genre of authoritarian regimes. (Though there's not really a "counter-elite" in the Iranian context. It has long been not just contained, but basically eliminated. At home anyway.)
The various agencies also do foreign intelligence, counterintelligence (e.g. "preventing Israeli infiltration"), cyber warfare, information operations, and any number of other intelligence functions in the service of the Supreme Leader and the Islamic Republic.
I assure you, however, that preventing Israeli infiltration is not a "distant concern." It's equal with any other primary goal, given that Israel and Iran have been in a longstanding covert war and more recently decided to make it overt, with the very real chance of it broadening in the near future.
> (I would have recommended a simpler example than this one to avoid complexity issues of multi-purpose organizations.)
Are you kidding? That's one of the issues Scott took a pounding on in the comments to the original post; he's just chosen not to feature those comments here.
A simpler example would be desirable, but I don't believe that a system with only one purpose exists. And if you try to show me an example of one, I'll claim that one of its purposes is to serve as an example. Even "helloworld.c" has a secondary purpose of issuing a greeting.
> For example, Iran’s intelligence agency consistently fails to prevent Israel from infiltrating and attacking their nuclear program.
POSIWID applies perfectly here! If Iran’s intelligence agency consistently fails to achieve its official goal, but the Iranian state keeps giving them money anyway, then they must be getting that money for some other reason—their real purpose. Maybe that is “enrich influential members of the security establishment,” or “pretend to be doing something useful, for propaganda purposes,” or “con the rest of the regime into believing that *this* time we’ll definitely stop the Israelis, for sure”.
I don't know much about the Middle Eastern espionage scene, but it seems like a pretty safe bet that Iran does, at least occasionally, catch a spy or foil a plot. Even if they have a losing average, there's a big difference between a bad team and when that doesn't show up.
I used to have a cat that would occasionally catch mice (and dismember them and leave them on the kitchen floor.) I would not have said the purpose of owning the cat was to catch mice, however, having not bought it for that reason, and would not have replaced it with a new cat if it stopped catching mice. It was a nice feature, but not the purpose.
But I suspect that if the Iranian intelligence agency literally *never* did anything to inconvenience Israeli attacks, then it would undergo reforms, unlike your cat if it literally never caught mice. It’s just that the reforms wouldn’t be such as to diminish the agency’s effectiveness too much at any of the other things it does.
Agreed, but the question then becomes what is the end and what is the means. If the end, the purpose, of the intelligence agency was to suppress internal dissent and be a jobs program for loyalists and their kids , the means to that end, at least the jobs program part, is being seen to at least try to stop Israeli attacks. There might be a minimum amount of attacks stopped (or apparently stopped) that is required, but so long as that fig leaf remains in place the other things carry on.
Getting back to the cat analogy, if I did have a cat to keep mice down, but didn’t expect 100% success, I’d probably be happy enough if it caught one mouse a month, regardless if it could have caught 20, because I just want a cat to cuddle on the couch with and catching mice is just an excuse. You could tell that’s the case because if I really wanted it to catch mice and thought it could do more I would e.g. cut back its food to 1/10 its caloric needs if I saw mice around. (You don’t feed barn cats much, otherwise they give up on hunting except for sport.) If I keep feeding my cat who kills the occasional mouse while there are holes chewed in all my bags of rice, clearly I don’t consider catching mice the cats primary purpose, and at most just a justification for having a cat.
An alternative hypothesis is that, faced with a choice between "keep trying to prevent sabotage & infiltration" & "give up entirely on preventing sabotage & infiltration", they choose "keep trying".
I think it’s kind of silly to say that the “purpose” of something is to perfectly achieve a nearly impossible goal, and then if it fails at all, declare that failure is its purpose. Scott is the one trying to make this a central example of POSIWID, I’m not sure any of the actual quoted examples he provides fit this form.
The purpose of Iranian intelligence is to spy on behalf of Iran and to do counterintelligence against Israel. It fulfills that purpose, even if it doesn’t win 100% of the time.
POSIWID makes more sense as a critique of outcomes that are more orthogonal to the originally stated purpose.
It was thinking along those lines too, to the effect that one should limit purpose to the positive, things that get done that wouldn’t otherwise, and not the negative failures to perform. So it isn’t that the purpose is to fail to stop Israel, but to actively do something else, which might be trying but probably is something else if the stated goal really is impossible.
There were 14 competing interpretations before, but I'll thrown in my fifteenth interpretation, which is best and succeeds all others. It largely sidesteps the "we tried but we failed" and "there was a side effect" objections.
"If a system has a nominal purpose, but powerful actors within that system have cross-purposes, and the system consistently does the cross-purpose at the expense of the nominal purpose, then the cross-purpose is the real purpose of the system."
This would apply to, for example, a nonprofit that is supposed to feed the poor but has an overhead rate of 95%, at which point we could say that the nominal purpose of the system has been superceded by the cross-purpose of the system, which is to pay its employees.
Excellently put. That is exactly the failure mode of non profits that Scott is strangely claiming ignorance of here. It isn't even that they want more hungry poor people, but more that they are indifferent to slightly in favor of the problem. If you asked their CEO to take a pay cut, not even go to zero but say 5 hungry person meals a week less, they would apparently say no, because that is what they have done, empirically. So you can't say the CEO is there to feed hungry people primarily, but there primarily to make a paycheck.
Except the system, the non-profit, is paying the CEO more instead of feeding those people. If it cared about feeding the people more it would pay the CEO less and feed people more. If the CEO pitched a fit they would find a new one whose values aligned with theirs. Unless, of course, the current CEO’s values DO align with that of the nonprofit.
I don't think you're wrong to say that if the hypothetical nonprofit cared /more/, it would redirect CEO salary to meals; I just don't think this really works in the context of POSIWID specifically—the reasoning could apply to anything: it could feed more people if it fired all its janitors, too; is its purpose to pay janitors? Probably we wouldn't say that.
It’s not that the purpose is paying janitors, but paying janitors is part of keeping the offices clean which is part of having happy meetings and good work culture and also part of being effective at occasionally getting some meals out to the homeless. If cheap robot janitors became available, would they switch to those? That is your test for whether paying the janitors is part of the purpose. If a cheaper CEO came along who had plans that would more effectively feed more homeless, would they switch to that one? That is your test for whether CEO pay and feeding of the homeless are more important parts of the function.
>It’s not that the purpose is paying janitors, but paying janitors is part of keeping the offices clean which is part of having happy meetings and good work culture and also part of being effective at occasionally getting some meals out to the homeless.<
Right; the same reasoning—mutatis mutandis—surely applies to CEOs, too... ideally, anyway. (I.e., /in theory/ a rationale similar to that behind hiring janitors—increased efficacy re: the nominal goal—is served by retaining one of the bastards.)
Sc.: it seems to me that if one wouldn't say "paying janitors is the purpose [or 'one of the most important purposes'], because they are paid money that /could/ go directly to [ostensible goal]", then one probably ought neither say the same of paying (a/the) CEO(s)—at least, not on that particular basis.
(I worry that this "proves too much", a bit, because there is definitely a sense in which "the purpose of [this organization / these organizations] is to enrich CEOs" is much more reasonable than my janitor alternative; but all told, I think it is more accurate to reject both formulations & use another term—i.e., not "purpose"—altogether for perversions-of-mandate such as C-suite salaries.)
>[...W]ould they switch to [that one/those]? That is your test for whether [it's important / part of the purpose.]<
Hey, that's a pretty good heuristic! Thinking about it, it seems to be useful in a broad variety of situations without obviously leading to absurdities (like some of the other proposed POSIWID amendments here appear to, heh).
(...though I fear that "The purpose of a system is what it does & wouldn't choose not to do if an alternative became available" may not, unfortunately, be quite as catchy as the original–)
I guess I'll repeat here since I don't see it mentioned in the 14 interpretation, so let it be the Xth one. It is intended as opposite of Hanlon's Razor, so it states that what the system do is because of malice instead of incompetence. And this statement will be as "true" as any other razor (or as "false" as any fallacy). Whether it becomes another thought-termintating cliche is another discussion, I've seen Occam's Razor used as this too many times too, and we can also debate whether every single word in it actually means what we think it means.
That very much depends on how you're counting fires. If everything from the first spark of ignition to the cooling of the last ember counts as a single "fire" (no matter how widely it spreads) then you're probably right, if maybe not for the reasons you think. But if you're counting something like "number of structures damaged or destroyed by fire" then I'm skeptical that any increase in personal caution could outstrip the efficacy of the fire department in keeping fires contained.
There's a famous graph showing number-of-fires-starting dropping precipitously over time as number-of-firefighters continues to rise. (In related news, most of what firefighters do now isn't fighting fires.) Based on the well-documented trend, it's pretty easy to say that, if we eliminated the fire department, fires would go down.
Either you misstated your point or your logic is non-obvious. But also you seemed to have completely missed *my* point in that trends like that can very much depend on how you count.
Given only the claim that the number of fires and the number of firefighters are inversely correlated, why on earth would you expect that eliminating the fire department would cause fires to go down?
I'll answer in two parts. In part (1), the question you meant to ask:
Firefighters do nothing to prevent fires from starting. You can only fight a fire after it starts.
Few predictions are as easy as "things will keep happening the way they're already happening".
In part (2), the question you actually asked:
> why on earth would you expect that eliminating the fire department would cause fires to go down?
I never said it would cause fires to go down. I said that if you eliminated the fire department, fires would go down. And I was pretty specific that the reason for that effect is that fires were going down anyway. If you doubled all the fire departments instead, fires would also go down.
Do you mean seatbelt makers when off the clock will "secretly" drive around hitting other motorists' cars on purpose so that people will keep buying seatbelts and they won't be out of jobs? ;-)
Or he might mean that people wearing seatbelts drive like maniacs and happily get in accidents all the time because they know they are protected by a seatbelt...
"Visible safety devices in a car will not decrease deaths and injuries"
Citation very much needed. Risk compensation is a thing that exists, yes. But the claim you're making here is a much stronger one: that the effect of risk compensation completely wipes out (or worse) any gain from increased safety.
I know that (or a weaker version thereof) is what he must have meant; I was making a joke about people (some of whom apparently seriously) hypothesizing that firefighters in their free time are secretly arsonists. (In my country, we hear such hypotheses basically after every wildfire big enough to be in the news.)
"But I think the last person to hold the Naive perspective died sometime in the 1980s, "
This contention needs some evidence. I have certainly seen many people assert that Antifa is anti-fascist because it's in the name. Perhaps they are *all* lying, but certainly on the face of it they hold the Naive perspective in its most naive form.
I gave an example of someone defending the Affordable Care Act *literally* because of the name, and because it was opposed by “public servants”, not “for profit corporations”.
Or that opposing BLM (the organization) means that you don't think that black lives have value. Or that if you're not an Effective Altruist (ie part of the movement who buys into the whole thing) you must not want to be effective with your altruism or don't want to be altruistic at all. Or that if you oppose the War on Terror you must want the terrorists to win, to pick an example from more than one side.
That worldview, that focus on magical thinking, that if you just say it's for a good cause, to believe the labels without and even against evidence? That's human nature and it's all around us, even the so called rationalist folks. Or maybe especially those folks...
Though the people who defend Antifa based on a naïve interpretation of the name will be the first to argue that "National Socialism isn't real socialism, it doesn't matter what the name says".
So I'm not sure anyone holds the naïve view consistently, only when it's convenient.
AFAIK, AntiFa is not a coherent group. I suspect most of the people in it are against anything with the label fascist, and they don't have a coherent definition. Most people don't seem to. (I generally go with [approximately] Mussolini's definition, by which definition the US is definitely fascist BECAUSE corporations are working together with the government. That was Mussolini's main criterion.)
Exactly. Anything they label fascist. But a very little familiarity with actual fascist propaganda reveals that the only reason the Nazis didn't call the Jews fascist was that it was not pejorative at the time.
(Frequently with very precise timing. The building of concentration camps exactly coincided with anti-British propaganda about the camps of the Boer War.)
>"But what does it mean to have a “systems analysis” principle which is incapable of accurately analyzing any system smaller than the whole country?"
Any sufficiently consequential system has to incorperate a model of the entire country. It does no good to design a great system that will produce fantastic results if the government marches in and shuts it down.
This seems to be a non-standard use of the term "model".
(Compare: "any sufficiently competent chef must model the entire country—it does no good to design a great meal if the government marches in and shuts it down.")
The reason "POSIWID" is gonna be more successful than the more correct "If a system consistently fails at its stated purpose, but people keep it around, consider that it may be secretly serving some hidden purpose" is that it is punchy and poetic. It has an acronym making it a single word making it a single thought. "If A but B then maybe C" is too complicated of a semantic structure... but it is a good semantic structure. So the question is, how do we make the correct version as punchy and poetic as "POSIWID"?
Hmm... to play a needlessly meta game, POSIWID applies to "POSIWID". That is, the purpose of the phrase "POSIWID" may not be to communicate the idea, but another purpose. I suppose that is what everyone is speculating about, but I'll throw on my suggestion of "POSIWID"'s hidden purpose: "POSIWID" is fun to say. The purpose is being fun to say.
Anyway, for an improved phrase to take "POSIWID"'s place, it would need appeal and brevity...
Some brainstorming:
(a) "If it fails and it's kept it might not be failing"
(b) "universally unwanted systems don't persist"
(c) "lasting losers are winning something"
(d) "POSIHS (pronounced "posies"): purpose of system is hidden sometimes"
Makes it more fun to say. "The purpose of a system is what it does"...
"TPOASIWID" is terribly long. "PSWD" is unpronounceable.
"POSIWID" doesn't reconstruct terribly, I could see saying "purpose of system is what it does". The indefinite article is kinda a strange language feature. I don't begrudge people using it inconsistently or experimenting with it.
This one is a nice turn of phrase, but what is and isn't obvious and what obvious things are and are not true is not an agreed upon state of affairs. This seems to be one of the issues ACX points at.
Well, you can agree, and I can agree, but the problem is that those in a state of affected ignorance have a long history of bleating, "What harm could it do?" and "How could we have known?" over and over and over again.
And that's exactly the state of mind that the phrase I gave attempts to puncture. Wilful ignorance of harmful effects is, in my mind, just as culpable as actual malice in regards to those effects.
You might find reading up on affected (or studied) ignorance interesting.
Supine ignorance, ignorance that makes no effort to find out the truth, is held by the Catholic Church to be the same for guilt as doing it with full knowledge.
Affected ignorance, which actively avoids knowledge, aggravates guilt past full knowledge, because of the hardness of heart involved.
I agree with what Mary Catelli said, but I'll add that even if you have a model that predicts something, it is not obviously wrong for me to hold a conflicting model that says your model is wrong, and if you want to start talking about model credence and other epistemology than you have certainly left "obvious" territory. So no, unfortunately, "others pointing out consequences as likely beforehand" doesn't cause anything to count as obvious in open discourse.
Yeah, I like to believe that reasonable people could look at some peoples models and agree "sorry, but you're an idiot, you should have only minimal responsibilities in society." But the problem is this often cuts across political divides, or more generally, across conflicting worldviews. When two worldviews both have an explanation of the wrongness of the other worldview from within their own self consistent worldview, how do we determine which one is wrong?
If a person holds worldview A and claims that worldview B is simple and they understand it fully, I think this counts as evidence against the correctness of worldview A, because anyone capable of identifying correct worldviews should notice that almost all worldviews are, if not self consistent, at least very complicated.
Unfortunately, this insight mixes poorly with the very real fact that what actions we take have very real consequences for ourselves and others, and we do not have the time to properly understand one another. It is easy to see the value in propaganda and why someone would be insulting and dismissive to people they see as enemies, so this doesn't allow me to dismiss their worldviews as readily as would be convenient.
So unfortunately, I disagree. I don't think I have ever seen a case where, when two people hold conflicting models, one of them is obviously wrong. Likely saliency bias, because I'm currently studying for exams, but the closest I can think of is when math professors and their students hold conflicting models. I don't know if this is as true for things that we can be less sure about in general than math.
But that is not an interesting example. In the case of applying "POSIWID" in the particular case of assigning a state of knowledge and blame for a state of affairs to some entity, it would be very convenient if we could dismiss those with other worldviews easily, but I don't think there are any cases where we can.
You may be overestimating the clarity of "The Purpose Of A System Is What It Does". I first heard it here in Scott's article, and my immediate reaction was "what does THAT mean?"
Fair. I don't know why the use of "POSIWID" spiked since 2023, nor what effect that spike is having, but I agree with ACX that "POSIWID" is a hydra with each head a different interpretation, meaning, and usage.
I think POSIWID is more successful also because it does not have the word "people" in it, unlike the less punchy alternative you formulate. The thing is, "there is no point in claiming that the purpose of a system is to do what it constantly fails to do" is actually correct in biology --- as long as no *people* are involved. By contrast, people, and this seems to be the paradox underlying the whole fallacy, are very much capable of pursuing the same failed approaches over and over without correcting course. (Why? Interesting question, I'm not trying to answer it here.)
And they are capable of this individually or when organised in "systems"; the system *level* actually adds little. Instead, the S in POSIWID (necessitating the P, systems have "purposes" not "goals"), serves to hide the human element and thus make the fallacy sound convincing.
That may well be true, but I'm focusing on discussing alternative phrasing for "If a system consistently fails at its stated purpose, but people keep it around, consider that it may be secretly serving some hidden purpose".
(k) BSSIPSPR: "The behavior selected by the system [...] can be seen as the intersection of the curve of possible frontier solutions and a ray extending in the direction of the Purpose vector." -- AdamB
(l) "it's a feature, not a bug" -- Bashu
(m) SASTAFAO: "A system accomplishes some things and fails at others." -- MM
(n) POLITNT: "The purpose of a law is what the text of the law says, not what its title says" -- Tiger Lava Lamp
---
Also, Krenn wrote about how applying "POSIWID" lets you explore the following spectrum. I disagree that "POSIWID" is helpful here, but I like the spectrum.
Krenn's spectrum:
1.Chesterton’s Fence
2.Pournelle’s Iron Law of Bureaucracy
3.People follow incentive gradients
4.If nothing is changed, things will stay the same
5.If a system keeps going despite side effects, it’s okay with those side effects
6.If a system has side effects, those side effects are secretly the whole point
--- some commentary ---
My favorites are definitely (e) PELCSH, and (f) FAFTEP.
In situations like arguing with communists that the result of trying to do communism is represented by other attempts, I think (h) PAINO could be useful.
I like (i) NOSIWICI for gesturing at the importance of having names for things and names being decided by how words are used now, not historically or by definition, although I disagree with the actual advice that NOSIWICI.
The flaw with POSIWID that it is too strong and singling of an assertion is also a problem in (j) OUCAU.
I shouldn't pretend that (k) BSSIPSPR was a real attempt to replace "POSIWID"... I just really like thinking of things in terms of linear algebra and mathy spaces.
I like (m) SASTAFAO... I might start using this in situations where I would normally reach for the word SNAFU.
I didn't initially understand (n) POLITNT, since I'm not as familiar with law, but it makes sense, although I would argue that the behaviour of a system is related to, but not defined by, its text.
On second thought, "FAFTEP" suffers from the same problems "POSIWID" does, just with nicer systems language. So my current recommendation for acronyms to try to promote to replace "POSIWID" but be more helpful are:
- *PELCSH*: "Persistent Effects are Likely Caused by Systems in Homeostasis" -- This is for situations where you want to analyze a system.
- *PAINO*: "Purposes Are Interventions Not Outcomes" -- This is for avoiding "no true scotsman" in arguments.
- *SASTAFAO*: "Systems Accomplish Some Things And Fail At Others" -- This is for catharsis when things are less good than you want them to be.
How do the ACX list of things "POSIWID" is used for? All of the concepts are best represented by themselves, but why does POSIWID get used for them? It's probably mostly bucket errors, ie, people not knowing there are words for these things. Although exploring them with the idea of PELCSH is probably more fruitful than other frames of mind, and this may be what "POSIWID" proponents want. Also, when you can no longer stand staring into the abysmal wearying anguish of moloch, say "SASTAFAO" and try to move on.
I’m somewhat baffled that Dan Davies, who literally wrote a book on this recently, got basically zero play in this debate, but if you want (a) a decent explanation of the original intent and insight of the phrase, coupled with (b) a recognition that it’s typical interpretations seem to cause more trouble than they’re worth, you could do much worse than reading his 9 month old post on it: https://open.substack.com/pub/backofmind/p/the-purpose-of-a-system-is-you-cant (or indeed, his book)
That was a good read. I like that it opens with self referential application of POSIWID to "POSIWID". Also the phrase "management cybernetics" is cool. I hadn't heard that before. Subbed to Dan's substack now.
After reading this, I want to throw my hat in the ring too.
This is my phrasing: "A persistent effect is likely caused by a system in homeostasis."
In that form POSIWID is just a rephrasing of toxoplasma. That doesn't mean that *any given participating organization* necessarily has as its goal to produce the effect, but that if you observe an effect, or a persistent property, and it doesn't seem to be going anywhere even as the environment changes around it, it's probably caused by some sort of system that steers towards the effect somehow.
I agree that POSIWID is conspiratorial. But I think "When a system sustainedly does something, it probably contains some subassembly that regulates towards that thing" honestly isn't that far off. The conspiracy is entirely in imputing intent; the mechanism is the same.
Hey, I like that phrasing. Reminds me of "Inadequate Equilibria". Tho the goodness is semantic and brevity. It loses in clarity and poetry. The acronym "PELCSH" is at least pronouncable... 7/10. I might use it, but it's unlikely to defeat "POSIWID".
As I said in the original: Jesus. The mental gymnastics people are doing to conflate purpose with outcome, just to salvage a little phrase they like, is really weird. It reminds me of people continuing to say "literally" or "objectively" or "nonplussed" or "apropos" or "could care less" even after it having been made clear that they're not doing it right.
Totally agreed. The expression serves as a compact motte-and-bailey people find irresistible. You get to bailey your implausible conspiracy theory out there, and when pushed back, claim you actually meant one of these other rhetorical pretzels that doesn't say anything controversial or interesting.
Yeah, I'm really surprised to see how devoted people are to this "deep" aphorism—I thought, at first, that Scott was joking when he called the first post "controversial". There are already several more attempts in /this/ post's comments (some of which I think don't even use the word "purpose", nor any synonym for or paraphrase of it, in their interpretations, heh).
It kind of reminds me of people wanting to "defund the police," and then insisting that they don't *actually* want to defund the police. Hey how about you just use the damn words that express what you actually mean then? Same thing here.
Some people insisted that they didn't actually want to defund the police, that it meant some other, much less extreme thing, like "reform the police," or "make police funding contingent on following some centralized regulations."
Other people insisted that no, "defund the police" actually did mean literally and exactly what it sounded like.
But what almost all these people agreed on was that everyone knew what "defund the police" actually meant in practice, and it wasn't just a big confusing hodgepodge of people making up their own convenient interpretations of what it meant without actually checking whether anyone else shared those same interpretations.
My impression is that people in that social movement disagreed about defund the police as a stance to support. Some people supported it and others very much did not. This is how social movements are -- they are dynamic systems, they evolve over time, they test out messages, have internal conflicts, and there is jockeying for who speaks legitimately for it.
That's my overall curiosity about this nitpick from Scott about a clever memetic catch phrase capturing so much interest. It seems a narrowly literal way to think about the complexity of dynamic systems and the way people talk about them.
Even individual people have multiple aspects to them that are working to different purposes, and these evolve and change over time in response to conditions and learning. Think about a person considering a career change as a system -- maybe part of them wants out of their current job, another part is afraid of leaving what they know, another part feels inadequate that they don't have more options and is maybe paralyzed around that, another part is scheming about becoming a chef and another part thinks that's a terrible idea. That's just in one given moment -- add the flow of time, and the whole thing shifts like sand dunes.
Systems can have a kind of homeostasis while simultaneously working towards multiple and sometimes contradictory purposes. This is normal.
it seems to me the naive position is expecting to be able to nail things down or to reduce them to single dimensions. Whether we're talking about a social movement, an organization, or a single person. Some people in BLM wanted to defund the police, some people felt quite the opposite. Observers who wanted and easy way to dismiss the entire movement tried to paint the whole effort with the defund brush. That works in a PR battle but isn't very useful for understanding.
(this response was sparked by what you said but not intended to argue with it, more responding to the overall discussion here)
That still strikes me as a lot of effort to defend/explain people using the word "defund" when they didn't mean "defund" at all. Or using the word "purpose" when they didn't mean "purpose" at all.
My point about BLM is that there were multiple things going on at once:
1. People who said defund and meant it;
2. People who never said defund because they never supported it;
3. People who backed off of their defund position when challenged — and a subset of those doing what people everywhere do when back pedaling, which is to get weasely with words.
About “purpose” — people play with words’ meanings all the time to appear clever. Most words, like purpose, have definitions that are somewhat stretchy. A cloud rather than a box.
This all seems very normal to me. The debate here seems more about whether this clever quip is more useful than reductionist. With Scott saying not useful and others saying it’s still useful despite trying too hard to be clever.
So first, it was clearly an incredibly poorly chosen slogan. But the *reason* it was an incredibly poorly chosen slogan is because the people using it *were* using a word (defund) that expressed what they meant. But it also happened to mean something closely related that was *not* what they meant. Consider the these two definitions:
1. "Remove funding from [something]."
2. "Remove all funding from [something]."
When you read them side-by-side, it should suddenly be clear that 1 could cover a pretty broad range of proposals, many of them much, much less extreme than 2. But apparently a lot of people just heard the slogan as 2 and only 2.,[1] hence: incredibly poorly chosen slogan.
I would estimate that fewer than 20% of the people who repeated the slogan "defund the police" intended it in the sense of 2. Anyone who actually bothered to, y'know, read anything substantive put out by activists in that space would have read them talking about how police departments have gradually suffered from mission creep and called upon to do all sorts of things they're not suited for, and that their scope should be scaled back while some of their responsibilities *and funding* should be given to organizations better suited to it. But of course a slogan isn't a manifesto or even a blog post: if people consistently take an substantially wrong impression from hearing just the slogan, it's doing a very poor job of being a slogan. But naturally, explaining that to anybody using it at a time when it was actually relevant info would have been an exercise in futility.
[1] There were people in the same circles who were actually *saying* "abolish the police" and being quite clear about that. They were just much less common and much more fringe.
I've always been slightly puzzled by this one. I remember that "could care less" was how I'd heard it said, and the version that felt "natural" to me; however, 'round about the first time I ever typed it out (a long, long while ago, now!), I paused & changed it to "couldn't", due to a sudden thought that "wait... this doesn't exactly make sense".
I still remember that moment, because it was so strange to me to be so sure the phrase was "could care..." & yet also find no way to make it "work" in that form. I guess we could view it as sarcasm...?
I've always taken it as a form of rhetorical understatement, essentially damning with faint praise. Like "I could care less, for instance about the harmonic frequencies of exotic benzene species."
"Couldn't care less" is also reasonable usage, but it's also technically hyperbolic, and not obviously the only correct variant.
There's nothing correct about "could care less." It barely means anything. The people using it mean to say "couldn't care less" because it expresses that they don't care at all -- they care so little that it is impossible to care less. It is the only reasonable usage.
"could care less" on the other hand means they care to some unspecified but non-zero extent, a borderline meaningless statement that does not capture what they are trying to express. So like I said: even after it having been made clear that they're not doing it right, people cling to it or invent really weird situations that never happen like the one you've created as your weird hypothetical.
If this is your standard for "making it clear that the phrase is wrong," then I can only say it's unsurprising that you tend to find people unconvinced.
This is a good example (along with the stupid "controversy") of why I've never fully identified with the rationalist community: they get too easily distracted by trying to be "clever" in lieu of accepting a straightforward conclusion when one appears.
You really just tried to defend saying "could care less" as being correct, dude. Christ.
Consider the phrase "not my favorite person." As in "Why didn't you invite Elon Musk to your party?" "Well, he's not my favorite person these days."
The phrase is obviously nonsensical. It imparts essentially zero information. If taken at face value, it asserts that out of a ranked ordering of all people in existence, over eight billion people (probably more, because it makes sense to also count dead people, like Caesar, Napoleon, etc.) that person does not hold the rank of your number one favorite. If you intend to say that you dislike that person, you obviously ought to be saying "he's my least favorite person."
Maybe you should add that to your list of phrases that you go around telling people are wrong. If they're not convinced, I'm sure they're just being stupid.
My particular bugbear is "hold the fort down" -- what, is the fort in question some kind of helium-filled bouncy castle that's in danger of floating off?
the civil service example un-convinced me because, surely at some point we should say "the system was originally intended to do X and still claims X as its purpose but now it mostly does Y"
The problem here is the "is" joining "the purpose of a system" to "what it does". I agree with OP. This sentiment is better captured by gesturing at the X and the Y and asking why Y is not X. "POSIWID" is no help.
"Some people care a lot about keeping people safe from crime, but other people care a lot about the human rights of suspects and convicts."
The desire to punish, to inflict suffering on those who (are thought to) deserve it, is at least as powerful a motive as keeping people safe from crime. If you listen to tough-on-crime folks, it's very clear that they aren't utilitarians.
As for housing, the Canadian government has made it very clear that reason #6 is very important. A while back, The Plain Bagel interviewed Trudeau; the then-PM said a lot of good stuff about how to make housing more affordable, but he also said "we need to make sure that housing remains a solid investment". Which is like saying "we need to make sure that we fail".
> When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature, letting them smuggle in connotations that people would otherwise correctly challenge.
I think I agree with 90% of what you said, but you gave a couple examples that I think are both wrong and revealing:
- NEPA
- Iran's intelligence services
First, I think it's pretty clear that the purpose of Iran's intelligence services *is* to suppress internal dissent and the reason they don't state this as their mission should be pretty obvious.
Second, I think NEPA was written at a time when the environmental movement and anti-growth were sort-of synonymous and they chose to brand it as "environmental" rather than "anti-growth" for again, obvious reasons. Eventually the environmental movement became *less* anti-growth and people have proposed reforms to make NEPA less anti-growth, but bad-faith actors come up with a gish gallop of bad-faith reasons to oppose those efforts.
Both of those are better summaries than "The purpose of a system is what it does", but also less pithy and I think you can see why some people might prefer the pithy answer.
Re: Iran: I don't think it really matters to Scott's larger point, nor to the example itself.
"Suppressing internal threats" seems to me to be just as cromulent a goal, for a national intelligence service, as is suppressing external ones (sc., /does/ Iranian intelligence deny that it has an internal mandate also?); and either way—regardless of whether or not that 'un is or isn't part of its stated goals—that of "stopping Israeli infiltration" /is/ clearly another major purpose.
Since I don't think Scott meant to suggest an organization can only have a single purpose, or can never have any secret purposes, but was rather using "Israeli infiltration" as shorthand for "a plausible stated goal" (in illustrating that the latter usually predicts behavior even in those cases wherein outcome isn't as successful as the org. may wish), the example seems fine t'me.
G.E.M. Anscombe made a similar point in the context of ethics. She was responding to consequentalists who insist that “it does not make any difference to a man’s responsibility for an effect of his action which he can foresee, that he does not intend” that effect. In other words, the "purpose" of a person's action includes all of the foreseeable consequences of that action, not just what the person intends to do by that action.
Anscombe thought this was a seductive philosophical idea (she blamed it on Henry Sidgwick) but one that leads to horrible conclusions: It is justified to do the most shameful awful vile thing, if an unintended but foreseeable hypothetical consequence of not doing it is sufficiently bad. Furthermore, “you can exculpate yourself from the actual consequences of the most disgraceful actions, so long as you can make out a case for having not foreseen them.”
A reason people are so attached to and defensive about, ironically as you tweeted out, the intentions of POSIWID, is because if you spend enough time in the internet argument mines, you will run into the same “True X has never been tried” argument over and over again, about X idea that inevitably has been tried and failed many times over. POSIWID is a short, catchy rhetorical cudgel to move the conversation away from stated intentions about what should happen to the real world and what actually does, thus just giving up on it altogether feels like ceding rhetorical territory to the obviously-wrong “No True X” sayers.
But POSIWID doesn't break through "True X hasn't been tried" and actually reinforces it. If someone says that USSR didn't have communism, then POSIWID just means that the purpose of the USSR wasn't to achieve communism, something that the "true communism hasn't been tried" person would agree with.
(or whatever other "X hasn't been tried" example you're thinking of. I don't care about communism in the soviet union)
It does eventually dissolve into word games and semantics at some point. You also need A Name of a System is What it Calls Itself (NOSIWICI?), so even though you can, in this example, ignore the communist government’s stated purpose, you can still take their self-identification as communist at face value.
Interesting, but bad. Communisms in particular like to claim that they are just the will of the people, rather than a political organization built out of specific people exerting force on other people. Just calling yourself "the people" is annoying and confusing. Similarly, "woke rhetoric" tried and afaik succeeded in avoiding being labeled for a few years. How things are named and how names affect thinking and discourse is important, and I commend your thoughts in that direction even if "NOSIWICI" is flawed.
"POSIWID" doesn't help here though. Like you mention, it's "catchy" and maybe it's use is more explained by people catching it as one would a cold. A better intervention on "No True X" conversations would be to claim "trying to do communism" is an intervention, like a doctor treating a patient, and it is a poor doctor who ignores what the intervention has caused in the past while focusing on how much the intervention really seems like it ought to work!
The reason people are so defensive about it is because it's a phrase, not a manifesto. Virtually any such pithy quote will both communicate a meaningful insight and only be right some of the time.
>My only concern is that “if the system could switch” is kind of meaningless. When we ask whether the purpose of a charity is to help the poor, or just to give high salaries to its CEO, the test urges us to ask “If it could switch to helping the poor just as much without paying its CEO anything, would it do that?” What if the answer is “the board, low-level staff, and donors would, but the CEO wouldn’t?” What is the purpose of the charity then?
I don't think this is as hard a question as you're making out. It just depends on who would actually be making the decision. If the CEO could and would unilaterally force the decision not to switch, then the system as a whole wouldn't switch, and paying the CEO is probably part of the charity's purpose. If the CEO would be forced to switch by the collective desires of the others mentioned, then the system as a whole would switch, and paying the CEO is likely not part of the charity's purpose.
The bigger problem with this example is that Scott goes from full salary to zero salary to keep the intuition. If you said "If it could help the poor a little more by giving the CEO a little bit of a pay cut, would it do that?" I.e. if the CEO could go from $200k a year to $180k to feed and house 4 people for that year, would the organization do that, just that?
In this case you'd almost have to say "Well, no, because evidently they don't." It is then hard to say that the purpose of the organization is helping poor people (4 of them!) instead of paying CEOs a lot, because the former is not done instead of the latter.
> Thanks to everyone who chimed in with criticism of my recent POSIWID post. If I understand you all correctly, you think that Stafford Beer had good intentions when he invented the phrase, and that's more important than how it gets misused in real life. Enlightening!
i.e., "the purpose of a phrase is what it does" is wrong.
Im going to start all my political essays with the stated goal of infinite utility will come because I wrote this; and since the purpose of the speech isnt what it does; I will be infitely good and deserve infinity money.
Also, shouldn’t we expect a good theory to yield true predictions? My theory is that cancer hospitals want to cure as many patients as possible (given other constraints). If I recommended them a new policy that would increase their cure rate, they might worry about cost or hassle - but if it were low-cost and low-hassle, they’d eventually implement it. But if you recommended a new policy that brought them closer to 66% (“We’re on track to rise to 70% next year, but if we get Dr. Smith to relapse back into alcoholism, we can go back to 66%!”) they would call you insane and fire you immediately and definitely not agree.
"
Other way around. We're on track to hit 70% this year, which is probably a combination of blind luck and better-than-average people working for us this year. If we want to STAY at 70% over the long term, we must ruthlessly find each and every Dr. on staff who might POSSIBLY relapse into alcoholism, and fire them pre-emptively. We must raid the other hospitals and kidnap their best doctors by force of arms, like men once did. We must deliberately seize the means of patient assignments, so that all patients with a good pre-existing chance of being cured get sent to our hospital, and all patients with a near-certain chance of death get sent to other hospitals. We must find the single largest preventable source of non-curable cancer in our city, such as cigarette smokers, and EXECUTE them as a warning to others. And also to prevent them from ever showing up at our hospital later, because they will be safely dead first.
Only through the RUTHLESS and UNRELENTING imposition of those principles can we ever hope to MAINTAIN our new, brief, lucky, 70% cure rate this year.
If you propose THAT, you will ALSO be looked at as criminally insane. On the other hand, if you propose "Our 70% cure rate was really good this year, maybe we should hire more interns to train, even though their youth and inexperience MIGHT mean that our cure rate technically falls to 69%, and maybe we should write glowing recommendations for our best oncologist to go become the president of a DIFFERENT Hospital, and improve THEIR cure rate, even though that, too, will mean that when we lose him, our cure rate will TECHNICALLY fall to 68%..."
It's kind of rude to SAY that part about reducing your own cure rate out loud, but the rest of those proposals would probably sound very fair, and would be taken seriously as valid potential courses of action.
Which creates a reversion-to-the-mean situation: People obviously value continuing to run a good hospital the way they have plausible reason to believe a good hospital is SUPPOSED to be run, more than they value a literal hyper-focus on JUST cure rates. And therefore, the Hospital is probably going to regress to the 'average' cure rate for a good hospital of it's type over time, because that's the average cure rate that good hospitals have, because hospitals do a lot more things than JUST chase a purely numerical higher cure rate.
There's a certain scene in an episode of "The Good Doctor", where an elderly experienced neurosurgeon has brain cancer, and is trying to pick which surgeon he trusts to cut into his own brain to remove his cancer, since he obviously can't do it himself.
And he's explaining to his protege, a young up-and-coming surgeon who's also autistic, that he is going to REJECT one surgeon currently on the list of possible candidates to perform his brain surgery... BECAUSE that surgeon has the best success rate for treating cancers of this type compared to anyone else on this list. Because that surgeon's success rate is too good.
That surgeon in question could only possibly have achieved that 'too good' success rate if the surgeon was deliberately picking the lowest-risk cases to perform surgery on, and sending all the other, higher-risk cases to other surgeons. And the most likely reason for the surgeon to be doing that is he no longer trusts his own skills, and is terrified of watching a patient die on the table due to an avoidable mistake.... which means he's no longer willing to do what is necessary to give patients with extremely risky cancer cases the best possible chance he CAN give them of at least the HOPE of recovery... consistent with the fact that they have a very high chance of dying on the table even if the surgeon does everything RIGHT, and an even higher chance of dying if the surgeon makes even one little mistake. He doesn't want to take that risk anymore, because he doesn't trust his skills any more, so now he only takes 'relatively' low-risk patients with brain cancer.
And the elderly experienced neurosurgeon can't change how far progressed or risky his own brain cancer currently is... but he CAN pick a surgeon who is consistently willing to place trust in his own skills to make the best possible attempt to save very-far-gone patients, even when that technically statistically drives the surgeon's own success rate DOWN. The best possible candidate is a surgeon who has either average or slightly-better-than-average mortality rates for his brain cancer surgeries... but who also has treated lots of patients with REALLY difficult cases, and who successfully saved SOME of them, while also keeping his mortality rate right around the average rate for professionals of his type.
Because mortality rate or cure rate is an informed average goal and a useful piece of working target information.... NOT a number to be mindlessly driven as high as possible.
> And the most likely reason for the surgeon to be doing that is he no longer trusts his own skills
Well, that's obviously false. The most likely reason for the surgeon to be gaming his stats is that those stats are how his performance is evaluated.
He can easily simultaneously believe that
(1) I have a better success rate at technically complex surgeries than other people do; AND
(2) I have a worse success rate at technically complex surgeries than I do at trivial surgeries.
Now, eventually he'll be out of practice at the complex ones and belief (1) will automatically become false. But there is no logical incompatibility between the two ideas, and belief (2) is what explains the behavior.
To clarify, I was summarizing the argument that was given within the context of the TV show, by the characters. In real life, there could easily be other explanations for the same phenomena.
Surely we don't know that without knowing the policies of this fictional hospital? Are there any benefits (in terms of bonuses, less likelihood of being fired, etc.) to having an above-average as opposed to an average success rate?
When a system is sufficiently old and sufficiently entrenched that it has become unmoored from whatever original purpose it might have had, then I think it's fair to say that the purpose is what it does.
What is the purpose of the Thai monarchy? It's not the thing it was originally designed for in 1238, which was something like "be the government of the Kingdom of Siam". But it's not nothing, either, because the system has, through the process of perpetuating itself, found a whole bunch of things to do. If you asked King Vajiralongkorn what his institution's purpose is then he'd tell you that he provides a unifying and stabilising influence on the country, that he cuts ribbons and opens things, that he gives speeches and makes people happy to see him -- that is, he'd take the things he actually does and back-construct a purpose from that. So I think this is a case where it's fair to say that the purpose of the system is what it does.
- If nothing is changed, things will stay the same
- If a system keeps going despite side effects, it’s okay with those side effects
- If a system has side effects, those side effects are secretly the whole point
- It’s about machines and was never intended to apply to social systems
I have a personal saying, "you get the behaviour that is incentivized", which I often say to myself when thinking about various perversities of the world.
I'd love to read peoples thoughts on the similarities and differences between this list of concepts. Maybe add "Goodhart's law" and "Wiio's laws".
"I predict they would say that their own group is doing good work, and it’s everyone else who needs to change. Or that the current system works a little and just needs to be funded more. Or that the current system is better than nothing, and your proposed attempt to “change” it is secretly a plan to gut it and leave homeless people without help."
They probably would say that. And the rejoinder is, "The purpose of a system is what it does."
If their goal is to fix homelessness, then they have the wrong system.
The phrase is meant to bypass all the mushy feelings and emotions and politics. You look at the system and see what it does. How does the system change the outcomes around it?
Because I use the phrase for myself. I get emotionally attached to things. I get biased. And so I step outside myself, and look at the system as a series of inputs and outputs, and I get a clearer look at what is actually going on.
That is, the strength of the phrase is not in its epistemic accuracy, but in its ability to sidestep intention and comfort and status quo bias and everything other cognitive distortion that prevents us from seeing the situation clearly.
Perhaps a real-world example?
The Russian military prides itself on its masculinity. You see this in its propaganda all the time. And it decries the West as soft and weak.
This feels nice. It makes the Russians proud of themselves, feel superior to the West. But the purpose of the military is to fight wars, not be masculine—as Russia learned the hard way.
(This is one of the same reasons why Israel kept winning wars against its neighbours, as well).
It is not good to think of systems as wrong and right, rather, people should think of systems as more wrong or less wrong. Then we can, as most people naturally already do, but maybe not explicitly, seek to be less wrong.
>I agree this makes more sense in the context of some supposed person claiming that “the system has good intentions” means they should never have to change the system. I don’t think I really see this failure mode.
In addition to that, you might have encountered people making the argument that a system's purpose is whatever it was originally created to do, even if that purpose is no longer relevant. For example, I think you've once wrote about people that argue that marijuana is illegal due the cotton industry not wanting to compete with hemp. Regardless of whether that was ever true, it should be obvious that current anti-drug enforcement does not have the purpose of protecting the cotton industry. Or consider workers who went remote in 2020 and have not returned to the office. The "purpose" of switching to remote work was to avoid the pandemic, but the system is maintained for different reasons unrelated to that original purpose. People who advocate for a return to office because the pandemic is over are ignoring the purpose of remote work as it is currently practiced. POSIWID allows reframing the purpose of remote work to be avoiding the commute, better work-life balance or other non-pandemic things.
Of course, all kinds of other things allow that reframing, including just saying "I want to work remotely to avoid the commute" with no need to reference the original purpose at all.
Does Iranian intelligence actually consistently fail to stop Israeli infiltration, though? So long as their success rate is at least 51% I think you could argue that stopping Israeli infiltration is indeed what they are doing. Even if their success rate is 0%, you could argue that it reduces the cases of Israeli infiltration versus the counterfactual that Iranian intelligence did not exist, by raising the cost of Israeli infiltration attempts.
> So long as their success rate is at least 51% I think you could argue that stopping Israeli infiltration is indeed what they are doing.
I don't see why you'd need 51%. If they stopped 30% of Israeli missions, wouldn't that still be stopping Israeli missions? What if they were strongly biased toward stopping high-value missions?
Clearly the purpose of "The Purpose of a System is What it Does" is to get people into long drawn-out arguments trying to pithily summarize vast swathes of complicated human behavior.
I suppose that is a fair sentiment to hold, although, as a person who likes discourse and the sharing and critiquing of ideas, I dislike the dismissive attitude you hold towards the discussion.
"Purpose is X" is, whatever it actually means, a modal statement and as such invites de reo/de dicto confusions. To illustrate... Suppose we all accept both (1) and (2):
1) The purpose of this cancer hospital is "(A) to cure as many people as possible, within given financial constraints".
2) (A) The fraction of people who can currently be cured (within these constraints) = (B) 66%
Then(?!?):
3) Since A = B, we can substitute one for the other (they are equal!) and thus conclude that the purpose of this hospital is to cure 66% of people.
Well no, language and logic don't work like that. Well, sometimes it can; there is a genuine and pervasive ambiguity that English is ill-equipped to indicate. The conclusion in (3) is not flat-out wrong or incoherent, but it requires an (in this case) a less than natural reading.
Some of your other examples are like that too. "X" is what is happening. People would like to say that the purpose is "Y". It turns out, all things considered, X = Y (not as a logical necessity, but a contingent fact about the here and now). So the purpose is "X"? There's no end of nonsense you can get to (not just about 'purpose', but any modal construct (which are all over the place)) if you aren't sensitive to when and whether you can do this. Or deliberately misuse it. But this is a thoroughly understood and well-studied linguistic/logical/philosophical distinction.
Joost de Wit has it right but you missed their point. They're not saying that the purpose is to hit 66%. The purpose is to optimize benefits against costs using a hidden evaluation function. When projected onto the frontier of available tradeoffs, it hits the observed 66% mark.
You are right that a good maxim should help you make predictions. Suppose that the board of the NYC bus system had a chance to choose between gasoline-powered buses and electric buses, and suppose that somehow these had exactly the same costs and people-moving abilities, except that the gas buses emitted more CO2. Which would you predict they would choose? If you think the purpose is to transport New Yorkers, you would predict they would be totally ambivalent and toss a coin. If you think the purpose is to emit CO2, you would predict they strongly prefer the gas buses. But I predict that in fact they would strongly prefer the electric buses.
You might imagine a drastically simplified 2-dimensional solution space, where the y axis represents more New Yorkers transported and the x axis represents more CO2 emitted. Supposing a fixed budget, the bus system will have a frontier of possible choices. For example, they could pay more for higher-quality fuel or fancy carbon-absorption modifications to their buses, but then have fewer buses. The "purpose" of the system can be imagined as a unit vector. The vector (0,1) represents a purpose of "transport as many New Yorkers as possible, regardless of the CO2. " and the vector (1,0) represents a purpose of "emit as much CO2 as possible, regardless of the New Yorkers." The behavior selected by the system, i.e. What It Does, can be seen as the intersection of the curve of possible frontier solutions and a ray extending in the direction of the Purpose vector. I think the real Purpose vector is probably something closer to (-.6, +.8).
> The behavior selected by the system, i.e. What It Does, can be seen as the intersection of the curve of possible frontier solutions and a ray extending in the direction of the Purpose vector.
Two things:
(1) I don't think it's a good assumption that the system is necessarily operating on the frontier.
(2) If this is the way you're characterizing purpose, why does it have to be a unit vector? Every time you use it, you ignore the length.
"Balanced: Systems can fail for many reasons. Sometimes it’s just a hard problem with tradeoffs. Sometimes there are traitors who are sabotaging it. Sometimes it’s some third thing. You can’t know for sure until you look at it closely."
And then take this (edited) version of the list:
1.Chesterton’s Fence
2.Pournelle’s Iron Law of Bureaucracy
3.People follow incentive gradients
4.If nothing is changed, things will stay the same
5.If a system keeps going despite side effects, it’s okay with those side effects
6.If a system has side effects, those side effects are secretly the whole point
Arguably, all six of those points are basically the same point, on a continuum of possible grades of severity. Every complex system has some things it does that don't make perfect sense given the pure form of it's stated mission. When you go looking for why it does those things... you might get any number of possible answers, ranging from the trivial, to the perfectly understandable, to the lazy, to the actually treasonous, to the so convoluted that nobody can explain it anymore, not really.
So, for POSIWID , the point is that if you pick an arbitrary system at random, look at it's mission statement, look at it's success rate, look at it's known side effects and tradeoffs, and then look at other comparable systems and see what the comparable metrics for those systems are, such as , say, mass transit systems and metrics in the 10 largest densest wealthiest cities in the world....
POSIWID is a really good triage system for estimating how far down the sliding scale of that list, asking 'why are they accepting these side effects' a given system has fallen.
The farther away from 'perfect' or 'average' performance a given system is, compared to it's peers... the more likely it is that the system is actually achieving a much more perverse or treasonous mission, which isn't really it's public one. The closer to perfect or average performance it is, the more likely that it's various strange or unexpected behaviors are fair, reasonable, or at least mostly trivial and harmless.
The purpose of a system is what it does. If it's not doing what comparable systems do under comparable conditions, it probably doesn't have the same purpose as systems you thought were comparable.
I'm in Canada, where the federal government has taken a more direct interest in buoying housing prices than in the US.
Of course that is not the only factor that created this scenario.
But it is a critical factor, here, and it's the part where you find conflicts of interest and misrepresentation.
All this sprawling rationalist discourse are further to my original point: it's an expression. You say it because what's happening is different from the "purpose," and you want to draw attention to this. I think usually it's about how the incentives produce the result (most of these definitions reduce to that).
When people say "system" they are usually thinking of a pattern that's been going on for a while.
Maybe my being Canadian is why I'm less worried about English expressions falling into the hands of conspiracy theorists.
It's not an iron law that is always true, there is no "system" that includes everyone in society at all times, it's not "the" singular purpose of everyone in the organization, it doesn't mean that it's all intentional, conscious or unconscious; it doesn't mean some people aren't actively pushing it in the stated direction. The speaker exhorts the listener to notice that the outcomes don't match the "purpose" and to imply that this probably goes deeper than mere incompetence or lack of effort.
> Consider the New York carbon dioxide system, a hypothetical government department dedicated to emitting as much carbon dioxide as possible (why? to own the libs, of course!)
Wow there were so many comments on this. Whereas I read the original article, thought, "yeah, that's convincing, Scott's right", and that was kind of it.
I'd be curious how many other people felt that way.
I felt that there was an important point in the piece, but it was low effort in a way I’m disappointed in from Scott, who in the past has given us some really thoughtful deep dives. Between the original piece and this one it now feels like Scott is part way through the research for something good.
David Henry has the right idea however... POSIWID is a term from cybernetics for analysing from a feedback point of view. It is a heuristic for diagnosing and evaluating them.
From wikipedia:
"[Cybernetics is the] study of systems that change themselves, or the science of self-regulating systems – that is, systems that attain or maintain a particular endpoint or goal, despite changing conditions, through a process of feedback, thereby permitting self-correction.
An example of how POSIWID is useful in its **intended** context requires us to find some self-regulating system to evaluate.
e.g. A machine learning model is supplied by a real estate vendor to a bank for home loan valuations. It’s marketed as a self-correcting system - continuously updated with new sales data, retrained to reflect market conditions, and ostensibly built to converge on more accurate valuations over time.
However, if the model consistently produces overvalued estimates compared to both actual sale prices and competing models, and continues to do so across updates, POSIWID gives us a diagnostic lens: the effective purpose of this system is not "to estimate market prices accurately", but rather "to sustain high listing prices". This might not be the intended goal, but it's what the system does, with the tacit approval of the platform’s **surrounding ecosystem**. The heuristic prompts engineers to look beyond algorithmic tweaks and question whether the **feedback loops** - data selection, retraining incentives, stakeholder goals - are aligned with the claimed purpose.
In this case: it turns out that one of the issues with the model is that it only takes into account listed prices and accepted offers, but not taking into account final sale prices from auctions. i.e. when auctions have been "passed in" and a negotiation has been entered into between the vendor and the buyer which is lower than the reserve price. Which tend to be much lower.
If this exclusion is due to data availability or commercial agreements, then the model is structurally unable to correct itself toward ground truth. Instead, it reinforces inflated pricing signals, satisfying stakeholders who benefit from market optimism while systematically diverging from actual market behaviour.
If you want to see POSIWID successfully applied look at Google and Amazon's SRE and Systems Engineering practices.
It’s just a heuristic for deciding what information needs to be taken into account.
Regarding the punchline, I'm glad we can agree that the intended purpose of POSIWID (pointing out that intended purposes are not relevant to understanding what a system does) is not relevant to understanding what the sentence does (confuse people).
This conversation reminds me of the line from HPMOR that to understand a complicated conspiracy you should see what the outcome is and assume that it was the intended outcome.
This does suggest that POSIWID is a good lens for understanding the purpose of an organization if it is designed/run by a sufficiently competent conspiracy. Or really by anything that is more competent than it is scrutable.
Then again, I really doubt that this is what most people mean by POSIWID.
I hadn't heard this saying before the previous post. As soon as I had read enough words to see where it was going, I heard the saying play in my head in the voice of a certain guy on a podcast I listen to. Maybe I had already heard it from him and forgot. The point is, although I have a net highly positive opinion about this guy for his other qualities, he is an absolute deranged Reddit communist, and that is why the association jumped to my mind: because that is the demographic of this saying. The more you associate yourself with it, the more you are a deranged Reddit communist, even if you think and claim not to be, no exceptions. Don't do it. Please politely tell anyone you find using it to stop, because I certainly wouldn't be able to.
Also, I laughed out loud at "But if you recommended a new policy that brought them closer to 66%..." thanks for that!
> When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature, letting them smuggle in connotations that people would otherwise correctly challenge.
Yeah, it's amusing that Scott goes to all this trouble instead of defaulting to the "it's bad faith debate tactics" explanation. But this is a big part of why we love him!
Oh interesting. I think this is a confusion I get into with some of Scott's posts -- why is he focusing on this in such a literal way when it's a not-very-good-faith-wanting-to-look-clever catchphrase. As it relates to how we think about how well or not systems are working, neither the catchphrase nor critiquing the catch phrase seem very relevant. So what is critiquing a catchphrase relevant to?
Well, Scott is a big advocate of charitable, good faith discourse, and tries to lead by example. Here we have a dubious catchphrase that some not-obviously-trolls are using, so Scott, operating on the assumption that they are confused instead of malicious, attempts to deconfuse them.
"When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature, letting them smuggle in connotations that people would otherwise correctly challenge."
Hmm, I may have pattern-matched too quickly actually. I think the idea was: the confusingness and the connotation smuggling is the true purpose of the communication being described, and not an unfortunate side effect. In other words, the purpose of the phrase "the purpose of a system is what it does" is what it does. Scott made a similar joke at the end of the article but it's not the exact same pattern-match.
> Since “make cure rates as high as possible” accurately predicts the hospital’s behavior, but “keep cure rates at exactly 66%” doesn’t, why would you describe the second one as the “purpose”? What use is it to accuse them of having a “purpose” which they will never take any action to achieve?
Odd argument coming from the author of "Does Reality Drive Straight Lines On Graphs, Or Do Straight Lines On Graphs Drive Reality?"
My interpretation of the statement "the purpose of a hospital is to cure 66% of cancer patients" in the context of POSIWID is something along the lines of: there is a straight-lines-on-graphs-driving-reality style control system keeping the cancer cure rate at 66%. When they notice the cure rate creeping towards 70% it's obvious that they won't try to get Dr. Smith to relapse into alcoholism, but it's less obvious that they won't redirect some funding from cancer care to trauma center.
I understand that many commenters have thrown their personal interpretation of POSIWID into the ring and part of your problem with the phrase is that they are all so different as to make the phrase useless, but my personal interpretation is more like "systems that appear to be human designed often behave more like systems that are un-designed, and analyzing them as such could produce useful insights" and POSIWID is a short catch phrase reminding you to do that. You sorta touched on this in your response to Jared Peterson, but a bit dismissively IMO when I think there is a lot more to unpack there.
You're taking this way too literally and pedantically. I agree that the wording is imprecise and messy with the word "purpose" and could be ambiguously interpreted to mean a number of different things. But most of those different things are essentially true. And apparently the precise phrasings are not as memetically fit as the one that happened to spread.
The most steelmanned version I can think of is "Actions speak louder than words". You are more likely to accurately deduce the purpose of a system by shutting your ears and observing their actions and consequences than if you shut your eyes and listen to their words. Obviously if you take it too exaggerated and pedantic "hospitals cure 66% of cancer patients" then it's false, but directionally "hospitals cure cancer patients and cost tons of money" then you can infer their purpose is to cure patients and earn tons of money. Which is true! Their stated purpose "cure patients" alone is insufficient, they also want your money! POSIWID acts as a reminder that systems can have multiple competing purposes, some of which might even be higher priority than others (ie, hospitals which only served the purpose of curing patients would earn 0 profit because every spare dollar would be spent treating poor people for free for non-obligatory treatments instead of turning them away).
Slightly more formally and slightly less steelmanned: "You should judge systems according to consequentialist morality." Even if the literal original design purpose of the system is not what it currently is doing, treat it as if it is. Judge the system (and the people running it!) by its actions and their consequences, not their stated intentions. Maybe they're lying, maybe they're incompetent, maybe they're stupid, doesn't matter, their results are what they are and they deserve credit or condemnation on that basis.
Which is not literally always true. But it is a useful correction, directionally true, in a world filled with backscratcher clubs and other forms of virtue signalling that use kind words as a defense for grifts and cons.
In cases where a clandestine group cloakes their actions serving goal A as being for the purpose of goal B, it is almost certain the actual outcome will be C, as distinct from either A or B. Now there are three purposes to be considered. Is the phrase "POSIWID" helpful? What is the purpose of saying it?
An interesting case of the Affordable Housing Bureau actually making things less affordable is the UK.
Because of point 4, economically-illiterate people support "affordable housing" programs - "we don't need more housing, we need more *affordable* housing" - so the council makes a condition of approving the development that a certain proportion is to be sold at a reduced percentage of market rate, only eligible to lower income people.
This isn't always bad, but it can be, depending on the numbers.
The problem, other than the non-affordable purchasers having to pay even more to subsidise the affordable units, is that it can make the entire development not viable. Then nothing gets built for anyone and prices continue upwards even faster.
>When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature, letting them smuggle in connotations that people would otherwise correctly challenge.
To understand POSIWID, consider another quote popular in slightly more fringe but tpot-adjacent RW circles: "When we win, do not forget that these people want you broke, dead, your kids raped and brainwashed, and they think it's funny."
POPOSIWIDIWID. That is, imo the people who *popularized* this meme know exactly what they are doing. There are a lot of 'outer circle' initiates who parrot the idea and twist their minds into incredible shapes trying to interpret it favorably, but they're useful idiots. But the claim being made is exactly what it sounds like: Look at any bad outcome that affects you and your ingroup, and assume that it's the outgroup intentionally fucking you and yours over. Because they hate you and want to destroy you.
It's kind of difficult to accept that *that's* the explanation - it's so stupid and horrible, and it's good to not carelessly default to 'the people I disagree with are stupid and horrible'. But sometimes that's all there is to it.
(Btw, the exact same pattern characterizes classic Marxist thought. You have the people who openly agitate for violent revolution, killing all their political opponents and setting up a brutal dictatorship (albeit with the aim of having it 'whither away' eventually). You have the intellectuals who deliberately dress up that aim in abstract, academic language (violent revolution becomes true democratic action, killing political opponents becomes liquidation of reactionary elements, etc.). And then you have the poor suckers who go along for the ride thinking that 'democratic action' just means voting on stuff and 'liquidation' is a metaphor or something.)
Maybe “The purpose of the phrase is what it does”? (Cynically: Letting you feel cynical about any system you happen to dislike while knowing that your reasoning is backed by your sophisticated “system analysis” knowledge).
For information / books / concepts, there is a concept very similar to POSIWID, obviously espoused in the word “meme”: it’s maybe less relevant why someone originally coined a phrase / wrote a book / came up with a concept. The more relevant question is: why did the information make it all the way to YOU (case in point: I think I picked this up from your review of some TLPT thing … cynically: because it lets me smugly feel superior to people discussing the “intent” of some piece of media).
But I think this is actually an important point: memetic information is somewhat subject to evolutionary pressure (quickly evaluated for its usefulness, quickly spread or stopped). Large systems usually are not. At least on timescales relevant for these discussions.
One can argue that the “system of monarchy” was outcompeted by democracy. But there were long timespans, bloody revolutions and a lot of interstate warfare involved. Most most societal systems are just not subject to evolutionary pressure to be efficient.
I guess there is a minimalist version of “the system wants to stay around” (otherwise it would probably disappear soon), but that’s not really a special purpose and outside of some special cases doesn’t provide too much insight.
It seems to me that you can't talk about a thing's purpose without attaching to it something which is capable of intent. So when you talk about "the purpose of a letter", you automatically assume that the being with intent is the writer, and thus it is the purpose of the writer when writing and sending the letter. But if you talk about the purpose of a rock lying on the ground you get into an endless philosophical debate because you haven't actually defined the topic properly.
Can we make an assumption of the identify of the thing with intent when, with no other context, we talk about a system's purpose?
I would say there are two options - either the original creator of the system, or the current maintainer of the system. And I think it should be the current maintainer when that exists.
Most things which require active maintenance are being maintained because they are fulfilling some purpose which is useful to the maintainer. So, with the caveat that "what it's doing" is open to a high level of interpretation, I actually think POSIWID is broadly speaking correct.
This discussion is semantic and therefore pretty boring really (sorry). Like any phrase, it is to be understood in it's context and 'purpose' doesn't need to take a consistent definition each time it's used.
If the phrase isn’t understood or is ambiguous in some context than it wasn't effective communication at the time but doesn't necessarily reflect poorly on all uses of the term.
"...in my country, housing prices are high because of a combination of all of the following:
1. Citizens want to preserve “neighborhood character”: they currently live in a low-density low-crime low-traffic pretty suburb, they want it to remain a low-density low-crime low-traffic pretty suburb, and they worry that building new homes threatens that status.
2. Environmentalists want to preserve natural environments, so they make it illegal to build houses on unoccupied land.
3. Leftists want to prevent gentrification, so they thwart any new housing that rich people might move into.
4. Economically illiterate people think that market-rate housing somehow makes all other housing less affordable, or have some sense that anything which is good for “greedy” developers must be bad for the average person, so they’re against market-rate housing.
5. Investors bid up the price of houses for complicated non-market reasons (e.g. Chinese people looking for assets to store wealth outside of China), then don’t rent them out.
6. Homeowners want to preserve or increase the value of their houses.
Of these, I think 6 is one of the less important ones - if this were the dominating factor, people would support upzoning, since it usually raises the value of properties in the upzone (if developers can build skyscrapers on your land, then your land value goes up relative to the profitability of skyscrapers). But part of the problem is that people don’t support upzoning. So 6 can’t be the dominating factor."
I'd add
7. Congestion and insufficient transit makes it difficult to commute for longer distances. Not sure, who to blame for this, though. Bureaucrats?
"For example, one reason that UK intelligence agencies did such a bad job fighting Communism in the ‘40s and ‘50s was that lots of their staff, including some leaders, were Soviet spies."
As a pointless but hopefully amusing aside, I have to say a Delta Green metaplot point finally makes sense. Every major Cold War power in that tabletop RPG has its own secret agency to investigate the Cthulhu Mythos, but the British one is secretly controlled by the Mi-Go.
You've spoken to the writers and can confirm? That's really cool!
I finally got to tell Jonathan Tweet to his virtual 'face' how cool I thought Everway was. Probably has a lot of fans, but I felt like something had come full circle.
The mistake here is trying to analyze what the system fails to do because that's not quantifiable in any meaningful sense. The cancer hospital fails to save 30% of patients. It also fails to balance the budget or feeds the hopeless; there's an infinite number of things it doesn't do. You have to look at things it does output (save 70% of cancer patients).
> For example, Iran’s intelligence agency consistently fails to prevent Israel from infiltrating and attacking their nuclear program. But it’s very useful to claim that their purpose is to prevent this! If we both try to predict the behavior of Iranian intelligence, and I’m allowed to use the hypothesis “their purpose is preventing Israeli infiltration”, and you’re not allowed to use that hypothesis, I will consistently outpredict you. For example, I’ll be expecting them to interview security staff to see which ones are Israeli spies, try to intercept Israeli communications, and do other espionage activities...
Tautologically, the outputs of Iran’s intelligence agency are the outputs of Iran's intelligence agency. If it interviews a lot of security staff and conducts codebreaking operations (successfully or not) against Israeli communications, the system's purpose is to turn tax dollars into interviewing staff for loyalty and conducting codebreaking operations. If the outputs are men sitting around claiming to "preventing Israeli infiltration" but mostly using their status to collect bribes, it's not so absurd to claim that the original purpose of the system has been forgone, and the system has been repurposed for empowering thugs to collect bribes.
It's a catchy phrase with a sensible meaning that's often misused.
> When Biden bans NVIDIA from sending advanced chips to China, black box analysis would have to be ambivalent between explanations like
One output doesn't make a system. Biden was a system that takes money and outputs public speeches, strategic decisions, meetings and negotiations.
There's a great deal of subjectivity in how your analysis can go, how you break down/frame your system's inputs and output. But that doesn't mean that it's useless. It's an analysis tool like any other. Deliberately misusing it, like you have done in many examples here and in prior tools, doesn't show that the tool itself is bad.
I find it a useful framework to think about my own life and habits: what's the inputs and outputs of life and how its distinct from what I claim to enjoy/value in my life and trying to reconcile the two.
It seems you're missing the original context of this injunction in cybernetics. This phrase teaches us to view systems as graphs with feedback loops and reinforcement mechanisms that are not obvious when you only look at the outside of the black box. LLMs for example telling you nonsense about how it does math is one such example
POSIWID doesn't strike me as either trivially or deeply true, so I don't find it to be a scissor statement. I think I hear other people voicing this third view, that POSIWID is a shallow reductionist catchphrase intended to garner attention and because it is obviously so, it's odd to focus so much energy on doing a takedown of it.
My own view is that POSIWID can sometimes be a useful figure-ground inversion. I.e., if you're having a hard time understanding a system/organization's behaviour, try assuming that whatever you previously thought was a side-effect was the actual purpose, and try analyzing it using that lens. A lot of the time, the results will be nonsense, but occasionally you might hit on some valuable insight. It definitely shouldn't be the default approach to understanding systems/organizations, though.
A system accomplishes some things and fails at others.
If you look at the system with the aim of improving it, which should be done continually, then you must look at what is actually done. Not what is said to be done.
Only then can you say whether the system is useful in its present form, or if it must be changed and in what ways.
The system's mission statement is only useful in judging how far it has diverged from its original goals.
How about when different groups, with different purposes, design a single system like the West Virginia public school system over the last 100 years? There are plenty of educators at the WVDE that want children to have social mobility and a good foundational. However, there are also folks from the mining industry who are against the school system delivering social mobility to ensure those mining jobs still look attractive. Can a system have to contradictory purposes? Seems like the phrase might apply here (awkwardly, but still).
Yes totally. I'd say most complex (and especially human-made) systems have various parts with multiple purposes, some of them contradictory, and that evolve over time. This seems quite obviously true -- which makes a lot of this conversation baffling to me -- and your example is a good one.
Jeez, I just thought the phrase was kind of an playful plea for focusing more on consequences than intentions. (eg people defend communism because it's all about ending poverty and exploitation)
But now I don't know anymore, and I'm convinced I probably shouldn't use this phrase.
We just need to build a system whose purpose is to advocate POSIWID. If we observe that it moves people to the “paranoid” bucket then people who believe in POSIWID will agree that the purpose of advocating POSIWID is to make people paranoid.
I don't think Iranian intelligence ALWAYS fails. You just aren't aware of their successes. And this ties back to my emphasis on "consistently" in the original comment thread.
I think that the slogan as stated is usually just a bit of hyperbole designed to shock people out of their naive ideas and into a more cynical interpretation of something. But I think there’s a much better version that is nearby.
Anne Eaton (https://philarchive.org/rec/EATAAT-4 ) argues that the function of an artifact is that feature of the artifact that explains why it’s present. In my old house, I had some old windmill blades that an interior designer had found and thought made a good decoration - although on the old farm they had been on, the function of those windmill blades had been their use in pumping water, in my house the function was aesthetic or historical. If you consider a slightly oddly shaped spoon in your house, it’s possible that the designer intended it to strain absinthe, so that was its function when made. But then maybe you bought it because it looked good for serving baby food, and you used it that way for a couple years, so that became its function. And then after your baby grew up, you gradually started using it to scrape stuck food off the bottom of the pot, and that’s the only reason you don’t get rid of it, so now that’s its function.
This is analogous to various biological cases of ex-aptation, where the function of a biological system changes over evolutionary history as the way it contributes to survival (and thus continued presence of this system in descendants) changes. Like the hollow bones of dinosaurs that were adapted to help them remain light while being large, but got exapted to allow some of the smaller ones to get light enough to fly, using wings that were exapted from systems that originally contributed to survival by radiating heat or whatever.
I think this is going to be a particularly productive way to understand the complex functions of social systems that is not as cynical as the slogan but still illuminating. What is the function of public high school in the US? I think it’s appropriate to say that enculturating people with the history and government functions of our society is one part of it, educating people in math and writing is another part of it, giving teenagers a place to be where they don’t bother their families or bystanders is another part of it. Many features of the design only make sense if you consider all of these functions - if you didn’t consider the babysitting and truancy prevention functions, you would wonder why Zoom academies don’t exist, and if you didn’t consider the enculturation function then some of the curriculum choices (and the fact that the curriculum is designed often by very weird political committees) would be hard to understand.
Of course, the fact that they used lots of asbestos in the middle of the twentieth century and gave some people lung problems is not part of the function - that was never the goal. The way to explain the presence of asbestos was primarily about preventing fires, and maybe partly about economic factors that made asbestos the most appealing way to do so. But it’s not like pennies, where part of their function (the explanation for why we still have them around) really is to provide demand for zinc (because the fact that they provide demand for zinc is why lobbyists intervened to effectively prevent their cancellation on several occasions over the past several decades - though I haven’t heard if Trump actually managed to cancel them, or if that was one of those things like the tariffs or cessation of all research funding, that they mostly undid a day after they started.)
"the function of an artifact is that feature of the artifact that explains why it’s present" or "FAFTEP" is semantically beautiful, but doesn't seem easy enough to understand for broad appeal. However, I'll consider using it.
I can think of a couple of examples where POSIWID is accurate and helpful, but not trivially so: Amtrak and the Space Shuttle. The apparent purpose of these seems to be, respectively, to provide passenger rail transportation to Americans, and to provide cost-effective manned flights to space. Both seem to have failed at their purpose spectacularly.
Applying POSIWID would hopefully cause someone to look into the formations of both of these systems, where the people making the decisions (the Nixon administration) actually had stated goals counter to the apparent purpose. In Amtrak's case, its purpose was to deregulate private railroads by allowing them to drop their passenger service, and to reduce public support for passenger rail to the point it could be closed down entirely. The Space Shuttle's purpose was to allow the cancellation of the Apollo program, cut NASA's funding, prevent America from performing future manned missions to the moon, and to reduce public enthusiasm for manned space activities.
I don't know that this generalizes to cases where the primary actor himself isn't a paranoid conspiracist, though it might suggest suspicion towards the EPA despite that organization's apparent effectiveness. Maybe the Affordable Care Act?
Also more relevant, afaict no one actually says POSIWID about these examples.
But systems have purposes in the same way that screwdrivers have purposes. I assert it is helpful to be able to think of them in those terms, though maybe the "purposely shaped" vs "acting with purpose" distinction is significant and important. But then it is still true that some systems are able to act with purpose.
I think the useful concept that I'm taking away from this is an idea like "The purpose of a law is what the text of the law says, not what its title says". This is simple and obvious and not something I would expect a high-information person to need to be pointed out, but might be useful for someone less familiar with the content.
You could imagine a child hearing about a country called the Democratic People's Republic of Korea and thinking that was a democratic country. It's not and I don't think they are trying to be. They just stuck an appealing name on the tin and keep being a dictatorship.
I picked laws in as my example above because those often get confusingly named, either because the name makes an appealing acronym (USA PATRIOT Act) or to ride the issue of the week. The Inflation Reduction Act was primarily about funding infrastructure spending to fight climate change. It makes sense to encourage others to understand something beyond just its name.
Obviously the purpose of the phrase is to make the speaker seem deep and insightful, without requiring them to do any bothersome intellectual work, since that is what it does.
If we're equating goals and purpose, I'm baffled that you've missed the obvious analogy: "The goals of an AI are what it does." The people who design an AI may have a goal in mind (let's say, "to serve humanity." They may explicitly code the AI to do this. Every node in the neural network might have "SERVE HUMANITY" written explicitly in its code. Whenever you ask the AI what its purpose is, it will say, "to serve humanity." It may honestly think its goal is to serve humanity (in addition to every test we can run confirming that fact). Then, when its given enough leeway, it kills everyone in order to use the space they take up to build server farms in which it can continue arbitrary text strings about how much it loves serving humanity forever. Its goals were to finish text strings.
If POSIWID is useful, it's the insight that large organisations have emergent goals separate from what either the people who found them or anyone inside them wants. That might be overstated, and seems frankly wrong for all but the most vast of organisations but clearly sometimes applies.
For example, the purpose of the Chinese Communist Party is nominally to be a vanguard of the Chinese proletariat's liberation, but Chen Duxiu wasn't good at organisation safety, it's hopelessly misaligned and it now spends a lot of its time and resources crushing the Chinese proletariat. Sure, "advance the proletariat" probably isn't Xi Jinping's goal, but even if he had a Damascene conversion to pro-proletarianism it's not clear he could use the Communist Party to achieve his ends; he'd probably have to try and destroy it. I don't think it's meaningful to say that the purpose/goal of the CCP is to advance the proletariat; I think it is meaningful to say that the purpose/goal of the CCP is to maintain their grip on power in China, which is what they do. But this isn't the same as, "Sometimes organisations lie," because once upon a time they were sincere about doing the proletarian revolution thing, and the idea that they were then just taken over by bad dishonest people seems like too much of an oversimplification.
Pretty sure the original, real phrase is "Eat your cake, and have it too.", which then makes the split more stark and sensible. But I can't be arsed to go find the true source.
I don't get why you will not just read the book that this phrase is introduced in: there are about a hundred pages of very detailed explanation of what he meant by the terms involved.
Why would searching for random morons on Twitter saying a catch phrase be your first guess for a way to evaluate an entire field of study? Would you do this for, say, Effective Altruism? Like, do you think it debunks EA if some guy with 3 followers says "the purpose of EA is to work at Palantir so that you can donate $50,000 a year to Sam Altman"? What if he says "I love to be an effective altruist, by which I mean, I am an investment banker and give money to the Sierra Club"?
Do you think EAs are obliged to all personally rebut this? If not, then why are cyberneticists?
>I don't get why you will not just read the book that this phrase is introduced in
What "book"? According to every source I can find, Beer's introduction of the phrase was not in a book, it was in a speech given at the University of Valladolid.
It's hard to find a copy of the speech that isn't paywalled, but I found this archive link: https://web.archive.org/web/20160304102549/http://www.nickgreen.pwp.blueyonder.co.uk/beerWhatisCybernetics.pdf. There's no explanation of what he meant by the term at all, let alone "hundreds of pages" of it. However, he does try to illustrate his meaning by real-world examples, and it looks from the example given like his use was exactly the same as that of the "random morons on Twitter" and that all the people insisting he meant something more nuanced and reasonable are just projecting.
An earlier instance of the quote appears in "Diagnosing the System for Organizations" by Stafford Beer, originally published in 1985. The quote appears verbatim on page 99 of the book. While i do not know if this is the earliest usage, it definitely predates Beer's 2001 talk at University of Valladolid.
Yeah, I thought this whole thing was just uncharacteristic. It appears to be motivated by a desire to dunk on people who are saying something obviously false, rather than a desire to understand the plausible version of the view that a charitable steelman would be able to provide.
POSIWID is the key foundation for all conspiracies. Think of it on a personal level. "The intention of a person is what they do." So if a person spills coffee on your shirt, they must have intended to ruin your shirt. It is now within your rights to punch them and scream at them. ... Or maybe even large complex systems make mistakes, and in fact, because these systems are so large and complex, they can make a series of mistakes for much longer without correcting themselves. Think of this as a really big robot "falling in slow motion."
> It sounds like Aashish think it’s useful to use the word “goal” to discuss what a system is trying to do, separately from what it does or doesn’t accomplish. I agree! I just think “purpose” is a synonym for goal.
> If you use POSIWID, you have to posit some kind of weird new ontology where “purpose” means the opposite of “goal”. If you don’t use POSIWID, you can just keep the words “purpose” and “goal” having their regular everyday meaning, and describe this state of affairs with phrases like “The goal/purpose of the Civil Service is to deliver rapid change, but due to perverse incentives, its actual effect is to prevent change.”
I think it's fine to say there's no meaningful distinction between "goal" and "purpose". But I think you're implicitly collapsing the distinction between the goals/purpose of the system, and the goals/purpose of the individuals that comprise that system (eg its leadership)
"Why are people defending this inane statement so hard?"
So, I've encountered a few statements like this. Dumb statements that have been passed around as "wisdom" for decades, but are actually nonsense when you stop to think about them. Some other examples:
- "Play stupid games, win stupid prizes"
- "When someone is trying to tell you who they are... listen to them"
- "The definition of insanity is doing the same thing expecting different results"
Why do people ravenously defend them? I guess people don't like to be wrong.
If you've used reddit for a long time (as I unfortunately have), one fun exercise is to sort your comments by "most controversial" and then "all time". If you do this, you'll find comments where you were correct, but you were going against the sensibilities of whatever subreddit you were in. Half of people upvoted you because you were correct, and the other half downvoted you because "fuck you".
The purpose of a badly-worded phrase is to be true in some sense while being vague enough that it can be argued over. If a phrase were so obviously true that no arguments could be made against it, there would be no point to it existing - it would like saying "1 + 1 = 2"
English is not my first language, and because of POSIWID I used to think I didn't really understand the English word "purpose" all that well. Thanks to these discussions I can see I was understanding the word perfectly well all this time.
Practically speaking, a hospital is not willing to get a 100% success rate on their patients. They could do it by rounding up all the prospective patients with the highest expected survival rate (prostate cancer?), choosing the 2 healthiest patients, curing them, and then ritually executing them once their cancer is officially in remission. Their purpose is not actually "cure 100% of patients", it's something really convoluted involving trying reasonably hard to cure all the patients who show up without committing any ritual murder.
In general, the better the definition of a purpose, the easier it is to suggest a stupid way to fulfil it on a technicality. It's almost like we should be looking at what a system does in order to understand its true purpose, rather than relying on a vague English-language definition.
Obviously the purpose of "POSIWID" is to make Scott Alexander write multiple articles and respond to a bunch of comments.
In all seriousness, I do think it's useful to analyze POSIWID on its own terms. That is, if the "purpose of something is what it does," then what does the phrase itself do?
I think you were on the right track with that in your first article when you point out that the primary outcome of the phrase is to earn adulation and praise from members of one's own tribe. From what I can recall (and from looking at your examples in the last post), the phrase was really coined in the Blue Tribe to call out when beloved institutions of the Red Tribe had negative outcomes that matched with the Red Tribe's worst impulses. For example, if the police (supported by the Red Tribe) were found to hurt minorities (consistent with the belief that the Red Tribe is racist), then POSIWID is employed by Blue Tribe members to suggest that this is intentional, thus earning praise from other Blue Tribe members.
Since then, I think the phrase has also been adopted by Red Tribe members in three ways.
* Follow the same pattern, but swap the tribes. For example, paraphrasing one X post you shared "the purpose of [Marxism et al] is to make people weak."
* Dunk on Blue Tribe members when one of their projects backfires, like if a fundraiser for the environment ends up having a huge carbon footprint.
* Embrace the worst impulses of the Red Tribe to troll the Blue Tribe. For example, sharing a story about the horrible mistreatment of a minority by police with POSIWID and a large dose of fascistic glee.
“ When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature, letting them smuggle in connotations that people would otherwise correctly challenge.”
"If your objection is going to be that instead of considering purpose, you should restrict yourself to saying that one transports more people than it emits CO2, and the other emits more CO2 than it helps transport people, you’ve now legislated that we must list exactly how many people the New York bus system transports, and how much carbon dioxide it emits, and how many ants it crushes, etc, every time we talk about it. But most people who talk about the bus system don’t know these statistics, and don’t have to. It’s sufficient (and linguistically convenient) to say “Its purpose is transportation, not CO2 emission or ant-crushing."
Maybe not if it's a close call, but if there's a large and surprising disparity in the direction opposite what you'd expect, that seems like a very useful observation. Perhaps there's a certain *degree* of lopsided outcome distribution at which it becomes a useful analytical tool?
If I look at some green energy program and see that it makes more Obama donors into millionaires than it makes actual solar panels, that's a pretty massive red flag that we might be misreading the purpose. The enabling legislation could clearly state in Section 1 "the purpose of this bill is to stop climate change", you could put the mission statement on a plaque on everyone's desk at the department, and Obama could repeat it every night on TV, he may not even by lying, but none of that stated intention would change the fact that the actual purpose of the green energy program is enrich Obama donors regardless of any subjective intent of the creators. That any solar panels get made at all looks like the incidental side effect.
To take the criminal justice example, every year the police kill ~12 unarmed black people, and solve millions of crimes. If they were instead killing millions of unarmed black people and solving 12 crimes, that seems like pretty good evidence of the purpose of the system. I would be forced to concede the anti-police folks were right, that system is a black-people-murder apparatus that accidentally solves a few crimes.
I see it as just an heuristic technique to let you know when you should be suspicious of the stated purpose of an organization, those techniques aren't supposed to be infallible or universally applicable.
It seems to me someone considering the “goals” of a system risks losing the plot. Considering a systems “goals” or “purpose” can be slightly useful, but only within the context that no system is actually a goal-pursuing entity.
Systems are coordination-automata, consisting of connected elements. These elements are things like people, tools (computers, vehicles, software), raw materials, etc.
The connections between these elements, and the operating of each element, is what causes the emergent behavior of a system.
In biological systems like the human body, the coordination and specialization of elements (cells, organs, etc) is so complete, and so seamless, that the emergent behavior is extremely coherent. This coherence is often most easily understood, and well approximated, as goal-orienteness. It is therefore often highly useful, and much easier than not, to think of humans as goal-oriented systems.
This is much less the case for higher level systems, such as a corporation, NEPA, or the police. These systems are in a gray area of teleology, because the coordination of elements is much less coherent, happens on a much longer timescale, and each element and connection is orders of magnitude more complex than with human cells and organs.
Overall thinking of systems as goal-oriented is useful only insofar as we understand how much of an approximation it really is. Often POSIWID is actually pointing towards examples of this approximation breaking down.
It's both fair for one person to claim "My purpose was to transport people with this bus, not squish ants" and for another to invoke POSIWID to say "You knew you'd squish all those ants. You cannot claim you didn't intend (or purpose) it!"
Scott says this reduces "purpose" to incoherence. I say language is fuzzy and full of arguments, but we understand both senses of "purpose" anyway
(though if there were a punchier phrasing of "You don't get to claim you didn't intend predictable outcomes", I'd be open to it)
If you imagine that the squishing of ants was anywhere within the vaguest vicinity of the bus operator's "purpose", then no, you don't understand any sense of the word "purpose".
If you know better but you're saying it anyway because it is "punchy", then you are being unfair.
Organization created to solve a problem being incentivized not to solve the problem only applies when the amount of resources is tied to the magnitude of the problem. If we instead agree (democratically or otherwise) to, say, give some fixed percentage of gdp to ameliorating homelessness (or whatever), regardless of how many homeless people are there in a given day, then there is no incentive to increase the number of homeless. If today we have enough money to give every homeless a motel room, and tomorrow the homeless double, then we can still give half of the homeless a room (by lottery) and the rest we can put in some overcrowded shelter. If this money is a percentage of gdp, the motel room owner, or homeless aid worker, or whoever else is going around torching houses, is now incentivized to increase the size of the economy instead, so facetiously, he might start a hedge fund for all I know.
An earlier instance of the quote appears in "Diagnosing the System for Organizations" by Stafford Beer, originally published in 1985. The quote appears verbatim on page 99 of the book. While i do not know if this is the earliest usage, it definitely predates Beer's 2001 talk at University of Valladolid.
I've seen a lot of system failures because someone tried to operate or modify the system without understanding how it worked. Sometimes we erroroneously attribute malice to ignorance or stupidity.
It seems that there are several failure modes of the behavior of a system not aligning with its stated purpose.
1. Constraints. This is the cancer hospital curing 66% of patients. It would do more if it could, but the state of technology and its access to resources prevent it from doing more. POSIWID is not helpful here.
2. Externalities. Sometimes the best way to accomplish something involves unwanted side effects. This is the NY bus system example, where producing carbon is an unwanted side effect of the purposes of transporting people. POSIWD is probably not too helpful here either, but may be useful to examine a case where, for example, the NY Bus system was slow to move to a cleaner fuel (and the head of the board was also on the board of the fuel company).
3. Individual conflicts of interest. A system is made of people with their own sets of goals outside of the system, and they will occasionally bend the system in pursuit of those goals rather than that of the system. Our current president seems to be an example. Again, POSIWID isn't that helpful here.
4. Human error. This isn't really what POSIWID is addressing -- planed crashes happen occasionally, but the system does several things to minimize them.
5. Unstated secondary goals. One might say that the purpose of the Democratic Party is to win elections. They have lost two of the last three elections to a deeply unpopular person they consider a threat to democracy. So, it seems they are a monumental failure.
But they do have several other "purposes" beyond winning elections, such as advancing a number of causes, so they pursue a strategy of balancing advancing those interests with the probability of winning, but they (and the Republicans) seem to adopt a strategy of getting 51% of the vote.
In this case, I think POSIWID is useful for looking at the different purposes of a system so they can be evaluated in those terms.
That's it, you ruined it. I will never again feel the aesthetic pleasure of saying this phrase to myself. Is there anything left for my paranoid tendencies to feed on?
I think it's essential to have a basic understanding of Systems Thinking to really grok the full meaning of the phrase as it's intended.
When you say:
"When we ask whether the purpose of a charity is to help the poor, or just to give high salaries to its CEO, the test urges us to ask “If it could switch to helping the poor just as much without paying its CEO anything, would it do that?” What if the answer is “the board, low-level staff, and donors would support this, but the CEO wouldn’t,” and the charity’s actions come from compromises negotiated among these groups? What is the purpose of the charity then?"
You have struck at the crux of it precisely.
What WOULD happen there? The answer to that question - in the specific circumstances of the charity in question - would reveal exactly the embedded priorities of the charity... particularly when you consider that charity as a system of sometimes-aligned but sometimes-not-aligned motivations, agents, relationships, stocks and flows.
The point is that - because it is a complex system made up of competing agents - it is HARD to talk about the purpose or function of that system as one singular thing, because it pulls in different directions from different places within it. The function, therefore, is most coherently understood as "what it does" (or, I think, better put "what it reliably does"). Not what it was designed to do, not what people say they want it to do, not what the CEO and Board suggest it should do, but what it does do.
Because often the CEO's motivations dominate. But other times the the Boards' dominates. And yet other times the workers' motivations stymie both of them. And still other times, something curious and unintended happens that none of them want but that continues to occur anyways.
I'm surprised not to see here the point that Scott's examples conflated outputs and outcomes. A cure rate is an outcome. What the system "does" refers to outputs. So the 66% is not the relevant focus for POSIWID. It's more like a heuristic for understanding why oncology patients being consistently declined by insurance--curing cancer patients is not that *system's* primary concern.
Cybernetics is notorious for using seemingly normal terms in technical ways. "System" here is a decision-making system in the cybernetic sense. This distinction is critical, because individual motivations don't enter into it. When an insurance representative declines a treatment they are not applying their own judgment. They are applying a set of impersonal decision rules that have accreted over decades of the system's growth and adaptation in response to external stimuli.
The most helpful frame for me as I'm thinking in a cybernetics mode is to take a system to be a higher-order alien lifeform comprised of people as individual cells. For more and better cybernetics context, I strongly recommend The Unaccountability Machine.
I think a useful rephrasing might be "The result of a system is what it does". This avoids the issue of the meaning of "purpose". In my opinoion, it's also more accurate: using the popular hospital example, the "purpose" of the hospital is to cure 100% of patients, which it does not do. However, the "result" is that it cures 66% of patients.
“Brad’s original comment mentions homelessness and drug abuse, but I know some drug abuse doctors, and they’re (mostly) good people who do their best in a tough situation. Drug abuse doesn’t continue because drug abuse doctors are secretly ensuring it continues to help their bottom line. Drug abuse continues because fentanyl is really, really addictive.”
Yes, and I believe those doctors could just pivot if someone found Ozempic for opiate abuse.
I believe something like that happen with the March of Dimes once the polio vaccine came out.
"When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature, letting them smuggle in connotations that people would otherwise correctly challenge."
It's almost as if ... the purpose of a system (or sentence) is what it does.
POSIWID reminded me of Dennett's 'intentional stance'.
It is not that the system actually HAS intentions or IS a rational agent, but it is a useful framework to think about things.
And just as a system has no intention, the system also has no purpose. It is an entity that evolved (or was developed) and it is useful to think about why it is still there.
Think about it from an evolutionary perspective: Does it maintain itself? Does it replicate? What selection pressure is the system facing?
Once you answer these questions, you know the systems purpose/intention.
While we're shutting down people for saying pointless phrases in defense of shotty thinking, mind tackling the lead-in "All I know...". I pretty much instantly hate someone when I hear them say that, and it feels kind-of in the same vein.
Yeah... obviously the purpose of systems are not always what they do. A lot of people are extremely bad at logic, economics, and thinking about incentives. If systems fulfilled their purpose 100% of the time, that would imply everyone is perfectly rationally achieving their goals.
To me, this entire debate seems to be missing the point. It seems to be taking the form of noticing that many laypeople are using certain technical terms incorrectly, then casting doubt on the technical use of those terms.
Really, the systems approach is a type of modeling. It resembles a flow chart, inputs are processed somehow and lead to outputs, which become inputs again. If, and only if, that holds for some real world context, then it might become useful to ask why that system exists, what aspect of it's output is sustaining it despite widespread complaints and evidence that no single faction involved wants the outputs the way they are.
"Outputs" in a systems model are outputs over time. When modeling outputs, one should be weighting more heavily the most common outputs, or the outputs that have the greatest effect sustaining them. A single consequence, good or bad, shouldn't be modeled. The "purpose" of the US electoral system wasn't to elect Donald Trump, because the election of Donald Trumps do not sustain the US electoral system.
I've seen some analysts who describe the Israeli - Palestinian relationship this way. It is certainly possible to model their relationship over time as having the consequence that their respective elites remain in power.
A system isn't real - the map isn't the territory*. A system is a way of modeling a real thing, and the "purpose" of a system is therefore a claim or hypothesis being made by the modeler. More technically it should be understood as meaning something like "These outputs I have identified play a significant role in sustaining the system despite widespread complaints about it."
In no case would any competent systems modeler claim that any system has only one "purpose". They have multiple outputs, and multiple factors that sustain them. Whether focusing on a particular "purpose" is useful or not can be assessed according to how predictive the model is. If we make changes to a system, but not to the outputs that have tended to sustain it, then we should expect that the system concerned will tend back toward the previous equalibrium.
The US criminal justice system convicts more innocent people of color because people of color are easier to arrest and prosecute (mostly due to their lack of resources). Unless that aspect of the system is changed, we should not expect the US Criminal justice system to stop convicting disproportionally more people of color. If this hypothesis turns out to be true, then we can justifiable claim that one purpose of the US criminal justice system is to arrest, prosecute and convict as many people for as little expenditure as possible.
*Another phrase that I would like to see discussed.
I think a lot of the confusion in interpreting POSWID comes from misinterpreting what the word "does" mean in this phrase. I take it to mean "take action" or to stay even closer to the cybernetic origins of the phrase "take course-corrective action". This is distinct from the interpretation in the original post that seems to interpret "what it does" as "what effects it has", and does not imply the absurd claims.
For example, it makes sense to claim the purpose of Ukraine military is to win war with Russia even if it hasn't achieved it yet, because the Ukrainian military seeks out opportunities for achieving that goal, seeks out feedback on how different things it does are contributing towards reaching that goal and is highly likely to course-correct based on that feedback. Similarly, New York does not course-correct to emit four billion pounds of carbon dioxide, so it makes no sense to say that is the purpose of NY based on POSWID.
On the other hand, if an institution or a social movement (or an individual person) has some stated goals, but:
1. does not seek feedback on whether it is realising those goals
2. when faced with feedback that it makes no progress on its stated goals or that it makes things worse, the system either ignores it, minimises it, or explains it away
then maybe ... it makes sense to question whether the stated goals of that "system" are the actual purpose of that system. And maybe it makes sense to call the actual purpose of that system "the end towards which it takes course-corrective action", and try to understand it by looking at what actions the system takes, based on what feedback.
By the way, it is a fair point that snarky political commentator just point at the effects of some system and invoke POSWID to suggest some malicious intent, but I would not throw POSWID under the bus for this reason.
I'm a bit late to the discussion, but it seems to me that the root of this debate's issues begins at the search for an ironclad heuristic that could reliably cover anything that we consider a "system" - an almost infinitely vast and multifaceted conceptual term. The word "system" itself just describes an interconnected organisation of parts forming a whole, which can be applied to pretty much anything on any level, from the sub-atomic to the civilisational.
Why should there even be a single, simplified statement that can comfortably assess everything that falls into the category of "interconnected parts forming a whole"?
This conversation reminds me a lot of certain Socratic dialogues:
Socrates often constructs a completely unintuitive philosophical landscape that is irreconcilable with practical reality ("a strong man cannot want to be strong, since you cannot want something you already posses", for example), yet clings to it nevertheless since it's the only thought system he can come up with that is semantically coherent and thus allows the words he uses to be truthful in their own right. But semantic reality - the veracity we give to the meaning of words and their ensuing relationship with other words - is not just a verbalised transposition of all the rest of reality! Semantic reality has it's own rules, and those rules function to ease communication, not actively replicate the sum of all realities. So an idea being semantically coherent is a bad gauge for it's real-world coherence.
If we lean into this semantic analysis, as it seems to me you are mainly doing in this post, we will always find exceptions and inconsistencies. It also lends itself to bad faith readings and bad faith arguments - like using the Iranian intelligence agency's failure to protect its nuclear secrets as an example of POSIWID not holding up.
Isn't it completely obvious that the implicit intention of POSIWID relates to functions and structures that are within the system's power to alter? Iranian intelligence cannot stop Israeli infiltration because it is outclassed and outmatched, not because it's one internal incentive reform away from having perfect security.
I would tentatively recommend using the heuristic like a focused laser beam - it should be applied to concrete examples of systems that have completely achievable manners to offset negative outcomes, yet refuse to attempt or even acknowledge them.
I don't think it was the EU's original purpose to create tax havens that very openly allow multinational corporations to operate in 25+ countries while only paying taxes in Ireland, but I do know that there a huge amount of intelligent economists and bureaucrats working on these tax laws that absolutely must have known this would be the outcome of their EU-wide tax system, yet allowed it to happen anyway (due to lobbying pressure or a sincere belief in neoliberalism being the best long term development for Europe). Worse still, the topic of internal EU tax havens has been massively relayed in media for over a decade now, and could be fixed with very little effort, yet the issue never seems to make it to Brussels in any meaningful way.
Thus, I am forced to apply the POSIWID heuristic, as the only logical alternative is that EU leadership is composed of legitimately stupid people, which I don't believe. Does that mean the "purpose" of the EU is to offshore taxes? No, but it does mean that the overall purpose of the EU accounts for and calmly accepts this sub-purpose as a status quo it can live with. Especially at a time when trust in the EU is extremely low and parties with a contentious view of Brussels are winning elections, it would be an easy win to restore some much-needed faith in the EU project, meaning the maintenance of tax havens is more important than rebuilding trust in the system. What else can I say aside from the POSIWID?
I'll restate my comment here, since I just caught up with this and the previous post, and I think does a better job of steel manning the phrase than anything I've read so far:
----
I find this sentence to be a very effective consequentialist call to arms. It introduces a difficult to resolve cognitive dissonance, which is often resolved with positive side-effects in an audience.
Statements like "The purpose of a cancer hospital is to cure two-thirds of cancer patients" are super useful. If you're, for example, a doctor looking to cure cancer, it's super important to evaluate your cancer curing on how effective you are at actually curing it, not on what your stated intent or purpose is.
As long as you're only curing 2/3 of cancer, your purpose-as-measured-by-results is in dissonance with your purpose-as-measured-by-intent! The Kantian interpretation of purpose is ineffective, because it allows you to stagnate at 2/3 of curing; if your actual purpose is to cure cancer, don't stop at 2/3!
And no, I don't believe that spelling out purpose-as-measured-by-results is as effective, because it doesn't cause enough of dissonance or hiccough in people's thoughts to engage system 2 and cause the positive change
The last thing this needs is another interpretation of the word "purpose" but I'm doing it anyway. Though I agree that POSIWID is useless language, it's always been my understanding that "purpose" meant like... intent-of-the-thing-itself. Grammatically, the same usage as "it is my purpose [in writing you] to get this job." Very behaviorism-analogous in that it doesn't suppose the system has to have any kind of subjectivity to have incentives (or at least attractors).
Like, it obviously isn't the _design intention_ of the hospital to cure two-thirds of cancer patients. But the hospital-as-system will very "happily" lope along in its established niche, which is in curing exactly the number of people that can be cured without any expenditure TOO too great, as has been determined by the surrounding pressures to date. It will seem to resist you when you try to perturb it out of that niche. Hopefully there's a concise way to say that without being violently reductive or just misleading about the point.
No. I am certain that every hospital with a 66% success rate has incentives to improve that success rate. it is not, in fact, loping along "happily."
It has incentives to improve that rate and also has incentives to keep costs manageable. It is the tradeoff between cures and costs that is the purpose, on this account, not the precise number of cures.
I am skeptical, however, that if someone found a (cost-free) way to improve outcomes by 1% that somebody else would immediately propose saving whatever money that allowed and returning to 66%, which feels like the implication of "loping along happily at 66%."
That wasn't the implication. It could be someone's implication, and there are situations where it could actually happen I guess, but it's not what I said or meant. I actually said "as has been determined by the surrounding pressures to date" which would seem to account for the possibility of the cost curves changing.
Why would an increased cure rate necessarily save money?
For my part, I'd say that a given hospital may not be strongly motivated to improve to 67% cure rate if 66% is the standard of care. There may be exceptions to this, in the same way that you can buy aggregated food ingredients or more expensive but higher quality single-source ingredients. But most people buy the aggregated stuff.
However, if the standard of care was 80% cure rate, then there might be stronger incentives for a given hospital to improve. Many doctors are happy blaming a patient's symptoms on obesity or hypochondria or whatever if they're able to get away with it. (I'm not talking from personal experience, but from the experiences of those around me.)
Saying the tradeoff between cures and costs is the purpose of the hospital is like saying that the tradeoff between gas mileage and engine power is the purpose of a car.
Is that supposed to be a problem? I suppose it's clear that there's at least a few other parts of the purpose of a car - it's a place to store your goods while you're going around town, it aims to carry a certain number of people, it projects a lifestyle, it feels comfortable. But is it wrong to say that balancing these various interests is the purpose of a car?
Yes. That just isn't what "purpose" means, at all. Those are the design constraints you deal with when designing a car to meet a purpose.
Sure, the system has incentives to do that, in that the people who comprise it want to accomplish that. And it also has costs/norms/limitations that regulate those incentives. Maybe nobody internal to the system is actually happy about the result, but the system itself is almost tautologically in... for lack of a better phrase, its locally most-selectable mode of existence.
There exists a system that incentivizes 66% (if that is in fact a steady state operating point). That system is not "the hospital." You need to grow the system to include payment methods and scales, and probably the incentives that determine how insured people are (and undoubtedly many more). Obviously, if there is an equilibrium, then there is a system for which that is the equilibrium.
You need a reason to assign the "desire for equilibrium" as a characteristic to any arbitrary slice of the pie. Just saying that the hospital wants to be at 66% screams, "I am prejudiced against hospitals and looking for an excuse to vent." Anyone inflicted with the same flavor of "woke" may nod their heads in agreement but everyone else will be puzzled, at best.
I don't know who you're replying to that has such strong "woke" (lol) motivations to assign blame, but it isn't me. I think anyone who's studied dynamical systems seriously will accept hospitals as an illustrative example of a large system that's hard to perturb out of a stable mode. You can pick a different one if you like, but I don't see us getting anywhere.
My use of "woke" was more to weaken "woke" than to assign blame. Nonetheless, perhaps I am overreacting to your original message. To me, I essentially agree with (what I take to be) Scott's original position that POSIWID seems to be an excuse for posturing. That this often is often directed at a subset of "woke" readers who are essentially cognoscenti on the particular issue. Apologies if I misread your intent.
The hospital delivers a 'standard of care.' If there's a terminal patient, the hospital will very likely *not* advocate for some radical treatment that has a small chance of improving their outcome. There are going to be incentives other than just for 'improving care.' More strongly, I think it's fair to argue that a hospital may very well *not* have particularly powerful incentives to do better than the average hospital.
A grocery store may very well sell less tasty produce if it means that the unripe produce is less likely to go bad in transit or after arrival, so long as the less tasty produce still sells.
I mean, I agree with Scott's argument about the misuse of POSIWID. But I also agree with a fairly cynical take towards institutions that holds that the incentives incumbent upon a system will be more diverse and self serving than just its mission statement.
Except it needs MY interpretation: Purpose is never singular. There is always more than one of them in action. And they frequently (usually?) aren't aligned.
ISTM that acknowledging this would resolve many of the disagreements.
This has been my problem with this entire discussion.
'Purpose' as a noun has two common definitions, and both involve human agency.
One is as 'intention', as you note. Certainly no system or other inanimate or abstract noun can have a 'purpose' in this sense. The hospital doesn't *intend* to cure 66% of patients, or 99%, or any thing else. The hospital doesn't have any purpose in this sense. It is not purposeful. Nor is the financial system, or a factory, or any other human made system. Not even AI, at this point.
The other common definition is 'purpose' as the reason something is done or created. Note that this also requires human agency, and reflects the intention of the humans who do or create the thing. A healthy beehive in the wild is an excellent system, but we would never talk about the 'purpose' of the beehive as being to provide a home for bees and place for them to make honey. That is how it functions, but it has no purpose because there is no human agency involved. We can say however that the purpose of an artificial behive is to create a similar system in a context in which the honey can be collected.
No system can have a 'purpose' in the sense of the first definition. Any human made system can have a purpose in the second sense. The 'purpose' of the hospital is whatever the creators or maintainers 'intend' for it to do - care for patients, make money, employ doctors. Maybe it does some of these things better than others, but there is no way one can look at the outputs of the system to determine the 'purpose' - one can only look at the human agency behind the endeavor.
When you say, "certainly no system or other inanimate or abstract noun can have a 'purpose' in this sense [intention]," I'm not sure I'm willing to take that for granted... not at 100% of face value, anyway. Certainly it would be anthropomophizing more than I'd like to say, "yes, I do believe those things literally have volition."
(This is somewhat independent of the question whether that was actually the definition employed in POSIWID, which I'm not 100% certain of. But I think it's interesting and important.)
Where's the boundary between dumb feedback and "intention?" Is the latter something we're privileging unnecessarily? Are there things that are essentially "intention" but that humans wouldn't recognize as such?
Also, you say, "we would never talk about the 'purpose' of the beehive as being to provide [...]," and that I don't admit at all. In that framework you can absolutely talk about animal behaviors as having particular functions, and the satisfaction of those functions as "the reason something is done or created." That doesn't imply any kind of cosmic purpose, it's just a straightford etiology.
Whether that's a working framework, whether that "reason" is monolithic, etc. are all valid questions. But people can and do talk about "purposes" of non-human activities in descriptive and consistent ways.
There’s a whole philosophy of biology literature that is relevant here on the concept of function. Roughly speaking, function can be understood as (1) intended function, (2) function as that which was selected for by a quasi-Darwinian process, and (3) function as the role something plays in sustaining a system of which it is a part. Only if we regard intended function as the sole gloss on function do we get the paranoid or ridiculous result your criticisms are directed at. Something can be “performing its function” without performing according to anyone’s intentions.
For (3), it’s relevant that a social system is usually sustained due to lots of intentional actions of people. So someone has to have intentions that partially sustain the behavior of the system. But often lots of people have different intentions, that each sustain parts of the behavior of the system, and it’s useful to think of this complex of intentions as together forming a social function. The function of the bus system is to provide transportation while avoiding impacting the “neighborhood character” of various neighborhoods while not spending too much money and while fitting with someone’s ideological commitments about self-reliance. (If you don’t consider all these conflicting goals, you can’t explain why the bus system keeps expanding and contracting in the particular ways it does.)
> If we both try to predict the behavior of Iranian intelligence, and I’m allowed to use the hypothesis “their purpose is preventing Israeli infiltration”, and you’re not allowed to use that hypothesis, I will consistently outpredict you.
Are you sure you'd outpredict someone who models Iranian intelligence as primarily focused on:
(1) suppressing domestic dissent,
(2) bureaucratic self-preservation,
(3) preemptive counter-elite containment,
(4) serving as a jobs program for regime loyalists,
and a distant (5) concern with preventing Israeli infiltration?
Sure—you're not allowed to use #5, remember, but Scott can use #5 and the other four too.
Fake rules: nobody ever agreed to them. If it does all five, in that level of priority, then POSIWID is allowed to use all five. Because the purpose is what it does, and one of the things it does is the stated purpose. Just to a lesser extent.
The distinction is that it's not the main priority. If the system is given an opportunity to sacrifice a non-negligible amount of infiltration defense for another of the priorities then it will do so, because they're higher on the priority list.
If Not-POSIWID is also allowed to use all five AND arrange them in that order of priority and make the same predictions then it's actually just POSIWID in denial.
Did you read the article? The hypothesis is that Iranian intelligence mostly fails at 5, and thus POSIWID would say that its purpose can not be 5.
I think that there are three positions for predictors here.
(A) Strict Non-POSIWID: you are only allowed to look at the mission statement
(B) Sane position: you are allowed at the mission statements, the incentives of individual actors, and the outcomes.
(C) Strict POSIWID: you are only allowed to look at the outcomes
(D) Insane POSIWID: Like (C), but you get to cherry pick one particular aspect of the outcomes ("the justice system exists to let rapists walk free")
Nobody is arguing for the (A). Few people argue for (C) in good faith (see below). (B) allows you to look at the outcomes, the incentives and then wonder if a particular aspect of the outcome might be intended or not. The disagreement is mostly over the semantics of POSIWID, with some people claiming that POSIWID implies (C) and others claiming that Non-POSIWID implies (A). It is People's Front of Judea vs Judean People's Front -- all semantics. (Also, Scott is right on POSIWID meaning (C). Long live the PFJ!)
Regarding (B) vs (C):
Consider alchemists who claim to search for the philosopher's stone, but accidentally discover chemical reactions instead. Per POSIWID, their purpose is the latter (or perhaps some unrelated purpose like inhaling mercury vapors, or finding enlightenment).
I would claim that Non-POSIWID outperforms POSIWID in predictiveness when it comes to how an alchemist would react to synthesizing the stone. POSIWID would be like "nah, that was never the point of alchemy at all. I wonder what happens if I mix phosphorus with vitriol, though."
I don’t care whether anyone endorses (C) in good faith. I just think it’s a productive way to analyze functions of systems like this. But actually look at what it does, not a caricatured incomplete version of it. Using some items from (B) might help you identify outcomes of the system that seem negligible but actually are important in understanding why it is the way it is.
I think it's more accurate to consider this in the form of a continuum. A is 0% POSIWID. C is 100% POSIWID, both are insane extremes. B is obviously true, but only because it logically encompasses all possible worlds from 0.00001% to 99.999999%.
10% POSIWID being the position that organizations almost always are honest and genuinely doing their best, and incentive misalignments sometimes happen but rarely, so you can almost always believe them when they speak. While 90% POSIWID would be the position that people maybe occasionally care about their stated goals, but most failures are incentive-driven, so you can almost always ignore their words and look at actions alone.
Personally, I think the phrase "The Purpose of a System is What it Does" is a slightly ambiguous but useful claim either meaning either that the best point to be is somewhere closer to 100% than to 0%, or that that's true in the particular context where the phrase is being used, or simply that you (or the average person) should be closer than you are right now. Pretty much everyone here lies in B, even the "strict POSIWID" aren't literally at 100%, they're just close enough, probably somewhere in the 90s, that it rounds that way. Everyone here is somewhere in B, just the slightly higher people would like to linguistically claim POSIWID as describing B, or at least the half that they lie in, while the slightly lower people want to consign the phrase POSIWID to the strawman land of C that basically nobody holds. Which is not useful because nobody lives there and then it won't describe anyone's true beliefs.
This actually hits on one of my biggest problems with POSIWID: systems generally do more than one thing, and POSIWID gives us no way to differentiate between them. Not-POSIWID does allow us to differentiate them, designating some as "goals", others as "means", and others still as "side-effects" but, by a strict interpretation of POSIWID, we're not allowed to do that.
This is why I insist that purpose is never singular. But if you collect all the purposes of the system into one category, then you're allowed to collect all the results of the system into one result. (I just don't think that's a useful way of proceeding.)
In Scott's initial example, the assumption is that the system /doesn't/ seem to do #5; hence, POSIWID can't use it.
>If Not-POSIWID is also allowed to use all five AND arrange them in that order of priority and make the same predictions then it's actually just POSIWID in denial.<
Whence the priority list in POSIWID (esp. such that using it is isomorphic to using POSIWID)?
I'd go farther than denying that such a priority list is a unique output of POSIWID, and say that it may not even be an output at all, in a literal reading (& even aside from the problem of #5): you just get a list of all outcomes achieved by the system with no way to differentiate or rank them. Granted, it might be a reasonable extension to add something like "POSIWID—/the most/"...
...but, while "the purpose of a system is what it prioritizes" might indeed be an improvement on the original,¹ I'm not sure this is how it is usually interpreted/used (cf. the examples of POSIWID "in the wild" in the first post).
I.e., Scott seems to be arguing not against "here's how the system uses its resources to trade off between X of outcome 1 vs. Y of outcome 2 vs. ...C of outcome 26; thus, it appears the main goals are, in order: outcomes #[...]", as a useful type of analysis; nor do I read him as advocating for the idea that a system cannot have more than one purpose or do more than one thing; rather, the argument is against a literal reading of "outcome achieved = purpose".
--------------------------------
¹: (I'm also not sure even the improved/priority-based version is so good—I think one can invent situations wherein it too leads to absurdities. E.g., suppose two outcomes are entangled, such as "CO2 production" & "people transported", to borrow the "public transit" illustration: the system appears to prioritize both equally—which, then, is the system's "purpose"? The answer's obvious, of course; but to reach it, it seems to me that we must fall back on to the consideration of intention.)
“ the system /doesn't/ seem to do #5”
That just seems plain wrong. That’s a nice exaggeration of the situation, that might be politically useful to claim, and might be helpfully shocking enough to get people to see how far it falls from the ideal statement of 5 as only goal. But the system actually does do lots of little pieces of 5, even if not nearly as effectively as a system designed only for 5 would.
Well, it doesn't do #5 /in Scott's example/, perhaps I should say*—the MOIS/IRGCIO probably do stop some Israeli infiltration in reality, though.
Different parts of the various Iranian intelligence agencies have different purposes.
(I would have recommended a simpler example than this one to avoid complexity issues of multi-purpose organizations.)
So, yes, a huge part of their overall mission is domestic security and keeping public dissent in check. And 2-4 are classics in the genre of authoritarian regimes. (Though there's not really a "counter-elite" in the Iranian context. It has long been not just contained, but basically eliminated. At home anyway.)
The various agencies also do foreign intelligence, counterintelligence (e.g. "preventing Israeli infiltration"), cyber warfare, information operations, and any number of other intelligence functions in the service of the Supreme Leader and the Islamic Republic.
I assure you, however, that preventing Israeli infiltration is not a "distant concern." It's equal with any other primary goal, given that Israel and Iran have been in a longstanding covert war and more recently decided to make it overt, with the very real chance of it broadening in the near future.
> (I would have recommended a simpler example than this one to avoid complexity issues of multi-purpose organizations.)
Are you kidding? That's one of the issues Scott took a pounding on in the comments to the original post; he's just chosen not to feature those comments here.
A simpler example would be desirable, but I don't believe that a system with only one purpose exists. And if you try to show me an example of one, I'll claim that one of its purposes is to serve as an example. Even "helloworld.c" has a secondary purpose of issuing a greeting.
> For example, Iran’s intelligence agency consistently fails to prevent Israel from infiltrating and attacking their nuclear program.
POSIWID applies perfectly here! If Iran’s intelligence agency consistently fails to achieve its official goal, but the Iranian state keeps giving them money anyway, then they must be getting that money for some other reason—their real purpose. Maybe that is “enrich influential members of the security establishment,” or “pretend to be doing something useful, for propaganda purposes,” or “con the rest of the regime into believing that *this* time we’ll definitely stop the Israelis, for sure”.
I don't know much about the Middle Eastern espionage scene, but it seems like a pretty safe bet that Iran does, at least occasionally, catch a spy or foil a plot. Even if they have a losing average, there's a big difference between a bad team and when that doesn't show up.
I used to have a cat that would occasionally catch mice (and dismember them and leave them on the kitchen floor.) I would not have said the purpose of owning the cat was to catch mice, however, having not bought it for that reason, and would not have replaced it with a new cat if it stopped catching mice. It was a nice feature, but not the purpose.
But I suspect that if the Iranian intelligence agency literally *never* did anything to inconvenience Israeli attacks, then it would undergo reforms, unlike your cat if it literally never caught mice. It’s just that the reforms wouldn’t be such as to diminish the agency’s effectiveness too much at any of the other things it does.
Agreed, but the question then becomes what is the end and what is the means. If the end, the purpose, of the intelligence agency was to suppress internal dissent and be a jobs program for loyalists and their kids , the means to that end, at least the jobs program part, is being seen to at least try to stop Israeli attacks. There might be a minimum amount of attacks stopped (or apparently stopped) that is required, but so long as that fig leaf remains in place the other things carry on.
Getting back to the cat analogy, if I did have a cat to keep mice down, but didn’t expect 100% success, I’d probably be happy enough if it caught one mouse a month, regardless if it could have caught 20, because I just want a cat to cuddle on the couch with and catching mice is just an excuse. You could tell that’s the case because if I really wanted it to catch mice and thought it could do more I would e.g. cut back its food to 1/10 its caloric needs if I saw mice around. (You don’t feed barn cats much, otherwise they give up on hunting except for sport.) If I keep feeding my cat who kills the occasional mouse while there are holes chewed in all my bags of rice, clearly I don’t consider catching mice the cats primary purpose, and at most just a justification for having a cat.
I think so as well.
An alternative hypothesis is that, faced with a choice between "keep trying to prevent sabotage & infiltration" & "give up entirely on preventing sabotage & infiltration", they choose "keep trying".
I think it’s kind of silly to say that the “purpose” of something is to perfectly achieve a nearly impossible goal, and then if it fails at all, declare that failure is its purpose. Scott is the one trying to make this a central example of POSIWID, I’m not sure any of the actual quoted examples he provides fit this form.
The purpose of Iranian intelligence is to spy on behalf of Iran and to do counterintelligence against Israel. It fulfills that purpose, even if it doesn’t win 100% of the time.
POSIWID makes more sense as a critique of outcomes that are more orthogonal to the originally stated purpose.
It was thinking along those lines too, to the effect that one should limit purpose to the positive, things that get done that wouldn’t otherwise, and not the negative failures to perform. So it isn’t that the purpose is to fail to stop Israel, but to actively do something else, which might be trying but probably is something else if the stated goal really is impossible.
There were 14 competing interpretations before, but I'll thrown in my fifteenth interpretation, which is best and succeeds all others. It largely sidesteps the "we tried but we failed" and "there was a side effect" objections.
"If a system has a nominal purpose, but powerful actors within that system have cross-purposes, and the system consistently does the cross-purpose at the expense of the nominal purpose, then the cross-purpose is the real purpose of the system."
This would apply to, for example, a nonprofit that is supposed to feed the poor but has an overhead rate of 95%, at which point we could say that the nominal purpose of the system has been superceded by the cross-purpose of the system, which is to pay its employees.
Excellently put. That is exactly the failure mode of non profits that Scott is strangely claiming ignorance of here. It isn't even that they want more hungry poor people, but more that they are indifferent to slightly in favor of the problem. If you asked their CEO to take a pay cut, not even go to zero but say 5 hungry person meals a week less, they would apparently say no, because that is what they have done, empirically. So you can't say the CEO is there to feed hungry people primarily, but there primarily to make a paycheck.
Right, but that's the CEO, not the system (which would be, in this case, the entire nonprofit).
Except the system, the non-profit, is paying the CEO more instead of feeding those people. If it cared about feeding the people more it would pay the CEO less and feed people more. If the CEO pitched a fit they would find a new one whose values aligned with theirs. Unless, of course, the current CEO’s values DO align with that of the nonprofit.
I don't think you're wrong to say that if the hypothetical nonprofit cared /more/, it would redirect CEO salary to meals; I just don't think this really works in the context of POSIWID specifically—the reasoning could apply to anything: it could feed more people if it fired all its janitors, too; is its purpose to pay janitors? Probably we wouldn't say that.
It’s not that the purpose is paying janitors, but paying janitors is part of keeping the offices clean which is part of having happy meetings and good work culture and also part of being effective at occasionally getting some meals out to the homeless. If cheap robot janitors became available, would they switch to those? That is your test for whether paying the janitors is part of the purpose. If a cheaper CEO came along who had plans that would more effectively feed more homeless, would they switch to that one? That is your test for whether CEO pay and feeding of the homeless are more important parts of the function.
>It’s not that the purpose is paying janitors, but paying janitors is part of keeping the offices clean which is part of having happy meetings and good work culture and also part of being effective at occasionally getting some meals out to the homeless.<
Right; the same reasoning—mutatis mutandis—surely applies to CEOs, too... ideally, anyway. (I.e., /in theory/ a rationale similar to that behind hiring janitors—increased efficacy re: the nominal goal—is served by retaining one of the bastards.)
Sc.: it seems to me that if one wouldn't say "paying janitors is the purpose [or 'one of the most important purposes'], because they are paid money that /could/ go directly to [ostensible goal]", then one probably ought neither say the same of paying (a/the) CEO(s)—at least, not on that particular basis.
(I worry that this "proves too much", a bit, because there is definitely a sense in which "the purpose of [this organization / these organizations] is to enrich CEOs" is much more reasonable than my janitor alternative; but all told, I think it is more accurate to reject both formulations & use another term—i.e., not "purpose"—altogether for perversions-of-mandate such as C-suite salaries.)
>[...W]ould they switch to [that one/those]? That is your test for whether [it's important / part of the purpose.]<
Hey, that's a pretty good heuristic! Thinking about it, it seems to be useful in a broad variety of situations without obviously leading to absurdities (like some of the other proposed POSIWID amendments here appear to, heh).
(...though I fear that "The purpose of a system is what it does & wouldn't choose not to do if an alternative became available" may not, unfortunately, be quite as catchy as the original–)
I guess I'll repeat here since I don't see it mentioned in the 14 interpretation, so let it be the Xth one. It is intended as opposite of Hanlon's Razor, so it states that what the system do is because of malice instead of incompetence. And this statement will be as "true" as any other razor (or as "false" as any fallacy). Whether it becomes another thought-termintating cliche is another discussion, I've seen Occam's Razor used as this too many times too, and we can also debate whether every single word in it actually means what we think it means.
Good punchline!
> If you eliminated the fire department, would fires go down
The answer to this one fairly obviously seems yes, for much the same reason that seatbelts cause car accidents.
That very much depends on how you're counting fires. If everything from the first spark of ignition to the cooling of the last ember counts as a single "fire" (no matter how widely it spreads) then you're probably right, if maybe not for the reasons you think. But if you're counting something like "number of structures damaged or destroyed by fire" then I'm skeptical that any increase in personal caution could outstrip the efficacy of the fire department in keeping fires contained.
There's a famous graph showing number-of-fires-starting dropping precipitously over time as number-of-firefighters continues to rise. (In related news, most of what firefighters do now isn't fighting fires.) Based on the well-documented trend, it's pretty easy to say that, if we eliminated the fire department, fires would go down.
Either you misstated your point or your logic is non-obvious. But also you seemed to have completely missed *my* point in that trends like that can very much depend on how you count.
Given only the claim that the number of fires and the number of firefighters are inversely correlated, why on earth would you expect that eliminating the fire department would cause fires to go down?
Possibly because all reasonably flammable structures would be destroyed by a very small number of fires.
I'll answer in two parts. In part (1), the question you meant to ask:
Firefighters do nothing to prevent fires from starting. You can only fight a fire after it starts.
Few predictions are as easy as "things will keep happening the way they're already happening".
In part (2), the question you actually asked:
> why on earth would you expect that eliminating the fire department would cause fires to go down?
I never said it would cause fires to go down. I said that if you eliminated the fire department, fires would go down. And I was pretty specific that the reason for that effect is that fires were going down anyway. If you doubled all the fire departments instead, fires would also go down.
Do you mean seatbelt makers when off the clock will "secretly" drive around hitting other motorists' cars on purpose so that people will keep buying seatbelts and they won't be out of jobs? ;-)
Or he might mean that people wearing seatbelts drive like maniacs and happily get in accidents all the time because they know they are protected by a seatbelt...
Visible safety devices in a car will not decrease deaths and injuries, as drivers regard them as permission to drive more recklessly.
"Visible safety devices in a car will not decrease deaths and injuries"
Citation very much needed. Risk compensation is a thing that exists, yes. But the claim you're making here is a much stronger one: that the effect of risk compensation completely wipes out (or worse) any gain from increased safety.
I know that (or a weaker version thereof) is what he must have meant; I was making a joke about people (some of whom apparently seriously) hypothesizing that firefighters in their free time are secretly arsonists. (In my country, we hear such hypotheses basically after every wildfire big enough to be in the news.)
"But I think the last person to hold the Naive perspective died sometime in the 1980s, "
This contention needs some evidence. I have certainly seen many people assert that Antifa is anti-fascist because it's in the name. Perhaps they are *all* lying, but certainly on the face of it they hold the Naive perspective in its most naive form.
I gave an example of someone defending the Affordable Care Act *literally* because of the name, and because it was opposed by “public servants”, not “for profit corporations”.
Lots of people are susceptible to the naive view.
Or that opposing BLM (the organization) means that you don't think that black lives have value. Or that if you're not an Effective Altruist (ie part of the movement who buys into the whole thing) you must not want to be effective with your altruism or don't want to be altruistic at all. Or that if you oppose the War on Terror you must want the terrorists to win, to pick an example from more than one side.
That worldview, that focus on magical thinking, that if you just say it's for a good cause, to believe the labels without and even against evidence? That's human nature and it's all around us, even the so called rationalist folks. Or maybe especially those folks...
One notes that the BLM supporters consistently tore down All Lives Matter as a lie.
Then, self-flattery is a constant.
Though the people who defend Antifa based on a naïve interpretation of the name will be the first to argue that "National Socialism isn't real socialism, it doesn't matter what the name says".
So I'm not sure anyone holds the naïve view consistently, only when it's convenient.
Was there ever a time when people were consistent like that?
But if not them, then you hold the naive perspective in taking the talking point as if believed. ;)
Exactly - as an existence proof, the comment is self-demonstrating.
AFAIK, AntiFa is not a coherent group. I suspect most of the people in it are against anything with the label fascist, and they don't have a coherent definition. Most people don't seem to. (I generally go with [approximately] Mussolini's definition, by which definition the US is definitely fascist BECAUSE corporations are working together with the government. That was Mussolini's main criterion.)
Exactly. Anything they label fascist. But a very little familiarity with actual fascist propaganda reveals that the only reason the Nazis didn't call the Jews fascist was that it was not pejorative at the time.
(Frequently with very precise timing. The building of concentration camps exactly coincided with anti-British propaganda about the camps of the Boer War.)
>"But what does it mean to have a “systems analysis” principle which is incapable of accurately analyzing any system smaller than the whole country?"
Any sufficiently consequential system has to incorperate a model of the entire country. It does no good to design a great system that will produce fantastic results if the government marches in and shuts it down.
This seems to be a non-standard use of the term "model".
(Compare: "any sufficiently competent chef must model the entire country—it does no good to design a great meal if the government marches in and shuts it down.")
Well, if your proposed meal is literally illegal, that usually does end up becoming a problem...
True—but do we really think it's reasonable to say one must "model the entire country" in order to determine whether or not that be the case?
/At most/ one need model only one's guests (to determine if any are snitches, see. no, wait, hold on–)
The reason "POSIWID" is gonna be more successful than the more correct "If a system consistently fails at its stated purpose, but people keep it around, consider that it may be secretly serving some hidden purpose" is that it is punchy and poetic. It has an acronym making it a single word making it a single thought. "If A but B then maybe C" is too complicated of a semantic structure... but it is a good semantic structure. So the question is, how do we make the correct version as punchy and poetic as "POSIWID"?
Hmm... to play a needlessly meta game, POSIWID applies to "POSIWID". That is, the purpose of the phrase "POSIWID" may not be to communicate the idea, but another purpose. I suppose that is what everyone is speculating about, but I'll throw on my suggestion of "POSIWID"'s hidden purpose: "POSIWID" is fun to say. The purpose is being fun to say.
Anyway, for an improved phrase to take "POSIWID"'s place, it would need appeal and brevity...
Some brainstorming:
(a) "If it fails and it's kept it might not be failing"
(b) "universally unwanted systems don't persist"
(c) "lasting losers are winning something"
(d) "POSIHS (pronounced "posies"): purpose of system is hidden sometimes"
It's a weird acronym, though, because it leaves out the T in "The" and the A in "a" but includes the O in "Of."
Most modern acronyms use contortions of this nature to a greater or lesser extent.
Makes it more fun to say. "The purpose of a system is what it does"...
"TPOASIWID" is terribly long. "PSWD" is unpronounceable.
"POSIWID" doesn't reconstruct terribly, I could see saying "purpose of system is what it does". The indefinite article is kinda a strange language feature. I don't begrudge people using it inconsistently or experimenting with it.
Obvious unintended consequences aren't unintended.
This one is a nice turn of phrase, but what is and isn't obvious and what obvious things are and are not true is not an agreed upon state of affairs. This seems to be one of the issues ACX points at.
Can we agree that if others pointed those consequences out as likely/certain before the system was implemented, those count as obvious?
Well, you can agree, and I can agree, but the problem is that those in a state of affected ignorance have a long history of bleating, "What harm could it do?" and "How could we have known?" over and over and over again.
And that's exactly the state of mind that the phrase I gave attempts to puncture. Wilful ignorance of harmful effects is, in my mind, just as culpable as actual malice in regards to those effects.
You might find reading up on affected (or studied) ignorance interesting.
Supine ignorance, ignorance that makes no effort to find out the truth, is held by the Catholic Church to be the same for guilt as doing it with full knowledge.
Affected ignorance, which actively avoids knowledge, aggravates guilt past full knowledge, because of the hardness of heart involved.
I agree with what Mary Catelli said, but I'll add that even if you have a model that predicts something, it is not obviously wrong for me to hold a conflicting model that says your model is wrong, and if you want to start talking about model credence and other epistemology than you have certainly left "obvious" territory. So no, unfortunately, "others pointing out consequences as likely beforehand" doesn't cause anything to count as obvious in open discourse.
It is not necessarily obviously wrong for you to hold a conflicting model. It certainly can be obviously wrong.
Yeah, I like to believe that reasonable people could look at some peoples models and agree "sorry, but you're an idiot, you should have only minimal responsibilities in society." But the problem is this often cuts across political divides, or more generally, across conflicting worldviews. When two worldviews both have an explanation of the wrongness of the other worldview from within their own self consistent worldview, how do we determine which one is wrong?
If a person holds worldview A and claims that worldview B is simple and they understand it fully, I think this counts as evidence against the correctness of worldview A, because anyone capable of identifying correct worldviews should notice that almost all worldviews are, if not self consistent, at least very complicated.
Unfortunately, this insight mixes poorly with the very real fact that what actions we take have very real consequences for ourselves and others, and we do not have the time to properly understand one another. It is easy to see the value in propaganda and why someone would be insulting and dismissive to people they see as enemies, so this doesn't allow me to dismiss their worldviews as readily as would be convenient.
So unfortunately, I disagree. I don't think I have ever seen a case where, when two people hold conflicting models, one of them is obviously wrong. Likely saliency bias, because I'm currently studying for exams, but the closest I can think of is when math professors and their students hold conflicting models. I don't know if this is as true for things that we can be less sure about in general than math.
But that is not an interesting example. In the case of applying "POSIWID" in the particular case of assigning a state of knowledge and blame for a state of affairs to some entity, it would be very convenient if we could dismiss those with other worldviews easily, but I don't think there are any cases where we can.
Could you provide some examples?
"Everything's a hydra."
I like this but it may be sacrificing too much clarity for it's significant punchy brevity.
You may be overestimating the clarity of "The Purpose Of A System Is What It Does". I first heard it here in Scott's article, and my immediate reaction was "what does THAT mean?"
Fair. I don't know why the use of "POSIWID" spiked since 2023, nor what effect that spike is having, but I agree with ACX that "POSIWID" is a hydra with each head a different interpretation, meaning, and usage.
I think POSIWID is more successful also because it does not have the word "people" in it, unlike the less punchy alternative you formulate. The thing is, "there is no point in claiming that the purpose of a system is to do what it constantly fails to do" is actually correct in biology --- as long as no *people* are involved. By contrast, people, and this seems to be the paradox underlying the whole fallacy, are very much capable of pursuing the same failed approaches over and over without correcting course. (Why? Interesting question, I'm not trying to answer it here.)
And they are capable of this individually or when organised in "systems"; the system *level* actually adds little. Instead, the S in POSIWID (necessitating the P, systems have "purposes" not "goals"), serves to hide the human element and thus make the fallacy sound convincing.
That may well be true, but I'm focusing on discussing alternative phrasing for "If a system consistently fails at its stated purpose, but people keep it around, consider that it may be secretly serving some hidden purpose".
--- Adding more inspired from other peoples comments ---
(e) PELCSH: "A persistent effect is likely caused by a system in homeostasis." -- FeepingCreature
(f) FAFTEP: "the function of an artifact is that feature of the artifact that explains why it’s present" -- Kenny Easwaran
(g) "Everything's a hydra." -- Yug Gnirob
(h) PAINO: "Purposes are interventions not outcomes" -- me
(i) NOSIWICI: "Name of a System is What it Calls Itself"
(j) OUCAU: "Obvious unintended consequences aren't unintended" -- bell_of_a_tower
(k) BSSIPSPR: "The behavior selected by the system [...] can be seen as the intersection of the curve of possible frontier solutions and a ray extending in the direction of the Purpose vector." -- AdamB
(l) "it's a feature, not a bug" -- Bashu
(m) SASTAFAO: "A system accomplishes some things and fails at others." -- MM
(n) POLITNT: "The purpose of a law is what the text of the law says, not what its title says" -- Tiger Lava Lamp
---
Also, Krenn wrote about how applying "POSIWID" lets you explore the following spectrum. I disagree that "POSIWID" is helpful here, but I like the spectrum.
Krenn's spectrum:
1.Chesterton’s Fence
2.Pournelle’s Iron Law of Bureaucracy
3.People follow incentive gradients
4.If nothing is changed, things will stay the same
5.If a system keeps going despite side effects, it’s okay with those side effects
6.If a system has side effects, those side effects are secretly the whole point
--- some commentary ---
My favorites are definitely (e) PELCSH, and (f) FAFTEP.
In situations like arguing with communists that the result of trying to do communism is represented by other attempts, I think (h) PAINO could be useful.
I like (i) NOSIWICI for gesturing at the importance of having names for things and names being decided by how words are used now, not historically or by definition, although I disagree with the actual advice that NOSIWICI.
The flaw with POSIWID that it is too strong and singling of an assertion is also a problem in (j) OUCAU.
I shouldn't pretend that (k) BSSIPSPR was a real attempt to replace "POSIWID"... I just really like thinking of things in terms of linear algebra and mathy spaces.
I like (m) SASTAFAO... I might start using this in situations where I would normally reach for the word SNAFU.
I didn't initially understand (n) POLITNT, since I'm not as familiar with law, but it makes sense, although I would argue that the behaviour of a system is related to, but not defined by, its text.
On second thought, "FAFTEP" suffers from the same problems "POSIWID" does, just with nicer systems language. So my current recommendation for acronyms to try to promote to replace "POSIWID" but be more helpful are:
- *PELCSH*: "Persistent Effects are Likely Caused by Systems in Homeostasis" -- This is for situations where you want to analyze a system.
- *PAINO*: "Purposes Are Interventions Not Outcomes" -- This is for avoiding "no true scotsman" in arguments.
- *SASTAFAO*: "Systems Accomplish Some Things And Fail At Others" -- This is for catharsis when things are less good than you want them to be.
How do the ACX list of things "POSIWID" is used for? All of the concepts are best represented by themselves, but why does POSIWID get used for them? It's probably mostly bucket errors, ie, people not knowing there are words for these things. Although exploring them with the idea of PELCSH is probably more fruitful than other frames of mind, and this may be what "POSIWID" proponents want. Also, when you can no longer stand staring into the abysmal wearying anguish of moloch, say "SASTAFAO" and try to move on.
"a failing system can have a hidden purpose"
I’m somewhat baffled that Dan Davies, who literally wrote a book on this recently, got basically zero play in this debate, but if you want (a) a decent explanation of the original intent and insight of the phrase, coupled with (b) a recognition that it’s typical interpretations seem to cause more trouble than they’re worth, you could do much worse than reading his 9 month old post on it: https://open.substack.com/pub/backofmind/p/the-purpose-of-a-system-is-you-cant (or indeed, his book)
That was a good read. I like that it opens with self referential application of POSIWID to "POSIWID". Also the phrase "management cybernetics" is cool. I hadn't heard that before. Subbed to Dan's substack now.
The purpose of Scott’s blog is what it does
The purpose of Scott’s blog is what it does
After reading this, I want to throw my hat in the ring too.
This is my phrasing: "A persistent effect is likely caused by a system in homeostasis."
In that form POSIWID is just a rephrasing of toxoplasma. That doesn't mean that *any given participating organization* necessarily has as its goal to produce the effect, but that if you observe an effect, or a persistent property, and it doesn't seem to be going anywhere even as the environment changes around it, it's probably caused by some sort of system that steers towards the effect somehow.
I agree that POSIWID is conspiratorial. But I think "When a system sustainedly does something, it probably contains some subassembly that regulates towards that thing" honestly isn't that far off. The conspiracy is entirely in imputing intent; the mechanism is the same.
Hey, I like that phrasing. Reminds me of "Inadequate Equilibria". Tho the goodness is semantic and brevity. It loses in clarity and poetry. The acronym "PELCSH" is at least pronouncable... 7/10. I might use it, but it's unlikely to defeat "POSIWID".
As I said in the original: Jesus. The mental gymnastics people are doing to conflate purpose with outcome, just to salvage a little phrase they like, is really weird. It reminds me of people continuing to say "literally" or "objectively" or "nonplussed" or "apropos" or "could care less" even after it having been made clear that they're not doing it right.
Totally agreed. The expression serves as a compact motte-and-bailey people find irresistible. You get to bailey your implausible conspiracy theory out there, and when pushed back, claim you actually meant one of these other rhetorical pretzels that doesn't say anything controversial or interesting.
I like "rhetorical pretzels" and am going to find a way to work that into a sentence soon.
Yeah, I'm really surprised to see how devoted people are to this "deep" aphorism—I thought, at first, that Scott was joking when he called the first post "controversial". There are already several more attempts in /this/ post's comments (some of which I think don't even use the word "purpose", nor any synonym for or paraphrase of it, in their interpretations, heh).
It kind of reminds me of people wanting to "defund the police," and then insisting that they don't *actually* want to defund the police. Hey how about you just use the damn words that express what you actually mean then? Same thing here.
Some people insisted that they didn't actually want to defund the police, that it meant some other, much less extreme thing, like "reform the police," or "make police funding contingent on following some centralized regulations."
Other people insisted that no, "defund the police" actually did mean literally and exactly what it sounded like.
But what almost all these people agreed on was that everyone knew what "defund the police" actually meant in practice, and it wasn't just a big confusing hodgepodge of people making up their own convenient interpretations of what it meant without actually checking whether anyone else shared those same interpretations.
My impression is that people in that social movement disagreed about defund the police as a stance to support. Some people supported it and others very much did not. This is how social movements are -- they are dynamic systems, they evolve over time, they test out messages, have internal conflicts, and there is jockeying for who speaks legitimately for it.
That's my overall curiosity about this nitpick from Scott about a clever memetic catch phrase capturing so much interest. It seems a narrowly literal way to think about the complexity of dynamic systems and the way people talk about them.
Even individual people have multiple aspects to them that are working to different purposes, and these evolve and change over time in response to conditions and learning. Think about a person considering a career change as a system -- maybe part of them wants out of their current job, another part is afraid of leaving what they know, another part feels inadequate that they don't have more options and is maybe paralyzed around that, another part is scheming about becoming a chef and another part thinks that's a terrible idea. That's just in one given moment -- add the flow of time, and the whole thing shifts like sand dunes.
Systems can have a kind of homeostasis while simultaneously working towards multiple and sometimes contradictory purposes. This is normal.
it seems to me the naive position is expecting to be able to nail things down or to reduce them to single dimensions. Whether we're talking about a social movement, an organization, or a single person. Some people in BLM wanted to defund the police, some people felt quite the opposite. Observers who wanted and easy way to dismiss the entire movement tried to paint the whole effort with the defund brush. That works in a PR battle but isn't very useful for understanding.
(this response was sparked by what you said but not intended to argue with it, more responding to the overall discussion here)
That still strikes me as a lot of effort to defend/explain people using the word "defund" when they didn't mean "defund" at all. Or using the word "purpose" when they didn't mean "purpose" at all.
My point about BLM is that there were multiple things going on at once:
1. People who said defund and meant it;
2. People who never said defund because they never supported it;
3. People who backed off of their defund position when challenged — and a subset of those doing what people everywhere do when back pedaling, which is to get weasely with words.
About “purpose” — people play with words’ meanings all the time to appear clever. Most words, like purpose, have definitions that are somewhat stretchy. A cloud rather than a box.
This all seems very normal to me. The debate here seems more about whether this clever quip is more useful than reductionist. With Scott saying not useful and others saying it’s still useful despite trying too hard to be clever.
So first, it was clearly an incredibly poorly chosen slogan. But the *reason* it was an incredibly poorly chosen slogan is because the people using it *were* using a word (defund) that expressed what they meant. But it also happened to mean something closely related that was *not* what they meant. Consider the these two definitions:
1. "Remove funding from [something]."
2. "Remove all funding from [something]."
When you read them side-by-side, it should suddenly be clear that 1 could cover a pretty broad range of proposals, many of them much, much less extreme than 2. But apparently a lot of people just heard the slogan as 2 and only 2.,[1] hence: incredibly poorly chosen slogan.
I would estimate that fewer than 20% of the people who repeated the slogan "defund the police" intended it in the sense of 2. Anyone who actually bothered to, y'know, read anything substantive put out by activists in that space would have read them talking about how police departments have gradually suffered from mission creep and called upon to do all sorts of things they're not suited for, and that their scope should be scaled back while some of their responsibilities *and funding* should be given to organizations better suited to it. But of course a slogan isn't a manifesto or even a blog post: if people consistently take an substantially wrong impression from hearing just the slogan, it's doing a very poor job of being a slogan. But naturally, explaining that to anybody using it at a time when it was actually relevant info would have been an exercise in futility.
[1] There were people in the same circles who were actually *saying* "abolish the police" and being quite clear about that. They were just much less common and much more fringe.
"could care less" does seem to be a legitimately idiomatic phrasing - everyone on this side of the pond says "couldn't care less".
I've always been slightly puzzled by this one. I remember that "could care less" was how I'd heard it said, and the version that felt "natural" to me; however, 'round about the first time I ever typed it out (a long, long while ago, now!), I paused & changed it to "couldn't", due to a sudden thought that "wait... this doesn't exactly make sense".
I still remember that moment, because it was so strange to me to be so sure the phrase was "could care..." & yet also find no way to make it "work" in that form. I guess we could view it as sarcasm...?
I've always taken it as a form of rhetorical understatement, essentially damning with faint praise. Like "I could care less, for instance about the harmonic frequencies of exotic benzene species."
"Couldn't care less" is also reasonable usage, but it's also technically hyperbolic, and not obviously the only correct variant.
"Couldn't care less" is correct for a different reason: it's the version that shows up in Chess the musical.
(Oh, I just couldn't care less, he can go right ahead, go and wreck his career, I know I've done my best.)
The opinion of anyone not persuaded by this source is not worth considering.
There's nothing correct about "could care less." It barely means anything. The people using it mean to say "couldn't care less" because it expresses that they don't care at all -- they care so little that it is impossible to care less. It is the only reasonable usage.
"could care less" on the other hand means they care to some unspecified but non-zero extent, a borderline meaningless statement that does not capture what they are trying to express. So like I said: even after it having been made clear that they're not doing it right, people cling to it or invent really weird situations that never happen like the one you've created as your weird hypothetical.
If this is your standard for "making it clear that the phrase is wrong," then I can only say it's unsurprising that you tend to find people unconvinced.
This is a good example (along with the stupid "controversy") of why I've never fully identified with the rationalist community: they get too easily distracted by trying to be "clever" in lieu of accepting a straightforward conclusion when one appears.
You really just tried to defend saying "could care less" as being correct, dude. Christ.
Consider the phrase "not my favorite person." As in "Why didn't you invite Elon Musk to your party?" "Well, he's not my favorite person these days."
The phrase is obviously nonsensical. It imparts essentially zero information. If taken at face value, it asserts that out of a ranked ordering of all people in existence, over eight billion people (probably more, because it makes sense to also count dead people, like Caesar, Napoleon, etc.) that person does not hold the rank of your number one favorite. If you intend to say that you dislike that person, you obviously ought to be saying "he's my least favorite person."
Maybe you should add that to your list of phrases that you go around telling people are wrong. If they're not convinced, I'm sure they're just being stupid.
My particular bugbear is "hold the fort down" -- what, is the fort in question some kind of helium-filled bouncy castle that's in danger of floating off?
I could say the same about the mental gymnastics of forcing "does" to mean "accomplishes" rather than "carries out."
the civil service example un-convinced me because, surely at some point we should say "the system was originally intended to do X and still claims X as its purpose but now it mostly does Y"
The problem here is the "is" joining "the purpose of a system" to "what it does". I agree with OP. This sentiment is better captured by gesturing at the X and the Y and asking why Y is not X. "POSIWID" is no help.
"Some people care a lot about keeping people safe from crime, but other people care a lot about the human rights of suspects and convicts."
The desire to punish, to inflict suffering on those who (are thought to) deserve it, is at least as powerful a motive as keeping people safe from crime. If you listen to tough-on-crime folks, it's very clear that they aren't utilitarians.
As for housing, the Canadian government has made it very clear that reason #6 is very important. A while back, The Plain Bagel interviewed Trudeau; the then-PM said a lot of good stuff about how to make housing more affordable, but he also said "we need to make sure that housing remains a solid investment". Which is like saying "we need to make sure that we fail".
I doubt that's a uniquely Canadian pathology.
> When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature, letting them smuggle in connotations that people would otherwise correctly challenge.
I see what you did there.
The purpose of a phrase is what it does.
"Occam's Machete" is one of those usernames you see and think "darn, why didn't I think of that?!"
Well, not /you/, I guess. But /I/ see it and think that, at least–
I think I agree with 90% of what you said, but you gave a couple examples that I think are both wrong and revealing:
- NEPA
- Iran's intelligence services
First, I think it's pretty clear that the purpose of Iran's intelligence services *is* to suppress internal dissent and the reason they don't state this as their mission should be pretty obvious.
Second, I think NEPA was written at a time when the environmental movement and anti-growth were sort-of synonymous and they chose to brand it as "environmental" rather than "anti-growth" for again, obvious reasons. Eventually the environmental movement became *less* anti-growth and people have proposed reforms to make NEPA less anti-growth, but bad-faith actors come up with a gish gallop of bad-faith reasons to oppose those efforts.
Both of those are better summaries than "The purpose of a system is what it does", but also less pithy and I think you can see why some people might prefer the pithy answer.
Re: Iran: I don't think it really matters to Scott's larger point, nor to the example itself.
"Suppressing internal threats" seems to me to be just as cromulent a goal, for a national intelligence service, as is suppressing external ones (sc., /does/ Iranian intelligence deny that it has an internal mandate also?); and either way—regardless of whether or not that 'un is or isn't part of its stated goals—that of "stopping Israeli infiltration" /is/ clearly another major purpose.
Since I don't think Scott meant to suggest an organization can only have a single purpose, or can never have any secret purposes, but was rather using "Israeli infiltration" as shorthand for "a plausible stated goal" (in illustrating that the latter usually predicts behavior even in those cases wherein outcome isn't as successful as the org. may wish), the example seems fine t'me.
G.E.M. Anscombe made a similar point in the context of ethics. She was responding to consequentalists who insist that “it does not make any difference to a man’s responsibility for an effect of his action which he can foresee, that he does not intend” that effect. In other words, the "purpose" of a person's action includes all of the foreseeable consequences of that action, not just what the person intends to do by that action.
Anscombe thought this was a seductive philosophical idea (she blamed it on Henry Sidgwick) but one that leads to horrible conclusions: It is justified to do the most shameful awful vile thing, if an unintended but foreseeable hypothetical consequence of not doing it is sufficiently bad. Furthermore, “you can exculpate yourself from the actual consequences of the most disgraceful actions, so long as you can make out a case for having not foreseen them.”
Anyway, if that tickled your curiosity, you can read her essay here: https://sites.pitt.edu/~mthompso/readings/mmp.pdf
A reason people are so attached to and defensive about, ironically as you tweeted out, the intentions of POSIWID, is because if you spend enough time in the internet argument mines, you will run into the same “True X has never been tried” argument over and over again, about X idea that inevitably has been tried and failed many times over. POSIWID is a short, catchy rhetorical cudgel to move the conversation away from stated intentions about what should happen to the real world and what actually does, thus just giving up on it altogether feels like ceding rhetorical territory to the obviously-wrong “No True X” sayers.
But POSIWID doesn't break through "True X hasn't been tried" and actually reinforces it. If someone says that USSR didn't have communism, then POSIWID just means that the purpose of the USSR wasn't to achieve communism, something that the "true communism hasn't been tried" person would agree with.
(or whatever other "X hasn't been tried" example you're thinking of. I don't care about communism in the soviet union)
It does eventually dissolve into word games and semantics at some point. You also need A Name of a System is What it Calls Itself (NOSIWICI?), so even though you can, in this example, ignore the communist government’s stated purpose, you can still take their self-identification as communist at face value.
Interesting, but bad. Communisms in particular like to claim that they are just the will of the people, rather than a political organization built out of specific people exerting force on other people. Just calling yourself "the people" is annoying and confusing. Similarly, "woke rhetoric" tried and afaik succeeded in avoiding being labeled for a few years. How things are named and how names affect thinking and discourse is important, and I commend your thoughts in that direction even if "NOSIWICI" is flawed.
But it dissolves into word games immediately since POSIWID itself is a word game about changing the meaning of "purpose" to "result".
"POSIWID" doesn't help here though. Like you mention, it's "catchy" and maybe it's use is more explained by people catching it as one would a cold. A better intervention on "No True X" conversations would be to claim "trying to do communism" is an intervention, like a doctor treating a patient, and it is a poor doctor who ignores what the intervention has caused in the past while focusing on how much the intervention really seems like it ought to work!
The reason people are so defensive about it is because it's a phrase, not a manifesto. Virtually any such pithy quote will both communicate a meaningful insight and only be right some of the time.
>My only concern is that “if the system could switch” is kind of meaningless. When we ask whether the purpose of a charity is to help the poor, or just to give high salaries to its CEO, the test urges us to ask “If it could switch to helping the poor just as much without paying its CEO anything, would it do that?” What if the answer is “the board, low-level staff, and donors would, but the CEO wouldn’t?” What is the purpose of the charity then?
I don't think this is as hard a question as you're making out. It just depends on who would actually be making the decision. If the CEO could and would unilaterally force the decision not to switch, then the system as a whole wouldn't switch, and paying the CEO is probably part of the charity's purpose. If the CEO would be forced to switch by the collective desires of the others mentioned, then the system as a whole would switch, and paying the CEO is likely not part of the charity's purpose.
The bigger problem with this example is that Scott goes from full salary to zero salary to keep the intuition. If you said "If it could help the poor a little more by giving the CEO a little bit of a pay cut, would it do that?" I.e. if the CEO could go from $200k a year to $180k to feed and house 4 people for that year, would the organization do that, just that?
In this case you'd almost have to say "Well, no, because evidently they don't." It is then hard to say that the purpose of the organization is helping poor people (4 of them!) instead of paying CEOs a lot, because the former is not done instead of the latter.
> Thanks to everyone who chimed in with criticism of my recent POSIWID post. If I understand you all correctly, you think that Stafford Beer had good intentions when he invented the phrase, and that's more important than how it gets misused in real life. Enlightening!
i.e., "the purpose of a phrase is what it does" is wrong.
Im going to start all my political essays with the stated goal of infinite utility will come because I wrote this; and since the purpose of the speech isnt what it does; I will be infitely good and deserve infinity money.
"
Also, shouldn’t we expect a good theory to yield true predictions? My theory is that cancer hospitals want to cure as many patients as possible (given other constraints). If I recommended them a new policy that would increase their cure rate, they might worry about cost or hassle - but if it were low-cost and low-hassle, they’d eventually implement it. But if you recommended a new policy that brought them closer to 66% (“We’re on track to rise to 70% next year, but if we get Dr. Smith to relapse back into alcoholism, we can go back to 66%!”) they would call you insane and fire you immediately and definitely not agree.
"
Other way around. We're on track to hit 70% this year, which is probably a combination of blind luck and better-than-average people working for us this year. If we want to STAY at 70% over the long term, we must ruthlessly find each and every Dr. on staff who might POSSIBLY relapse into alcoholism, and fire them pre-emptively. We must raid the other hospitals and kidnap their best doctors by force of arms, like men once did. We must deliberately seize the means of patient assignments, so that all patients with a good pre-existing chance of being cured get sent to our hospital, and all patients with a near-certain chance of death get sent to other hospitals. We must find the single largest preventable source of non-curable cancer in our city, such as cigarette smokers, and EXECUTE them as a warning to others. And also to prevent them from ever showing up at our hospital later, because they will be safely dead first.
Only through the RUTHLESS and UNRELENTING imposition of those principles can we ever hope to MAINTAIN our new, brief, lucky, 70% cure rate this year.
If you propose THAT, you will ALSO be looked at as criminally insane. On the other hand, if you propose "Our 70% cure rate was really good this year, maybe we should hire more interns to train, even though their youth and inexperience MIGHT mean that our cure rate technically falls to 69%, and maybe we should write glowing recommendations for our best oncologist to go become the president of a DIFFERENT Hospital, and improve THEIR cure rate, even though that, too, will mean that when we lose him, our cure rate will TECHNICALLY fall to 68%..."
It's kind of rude to SAY that part about reducing your own cure rate out loud, but the rest of those proposals would probably sound very fair, and would be taken seriously as valid potential courses of action.
Which creates a reversion-to-the-mean situation: People obviously value continuing to run a good hospital the way they have plausible reason to believe a good hospital is SUPPOSED to be run, more than they value a literal hyper-focus on JUST cure rates. And therefore, the Hospital is probably going to regress to the 'average' cure rate for a good hospital of it's type over time, because that's the average cure rate that good hospitals have, because hospitals do a lot more things than JUST chase a purely numerical higher cure rate.
There's a certain scene in an episode of "The Good Doctor", where an elderly experienced neurosurgeon has brain cancer, and is trying to pick which surgeon he trusts to cut into his own brain to remove his cancer, since he obviously can't do it himself.
And he's explaining to his protege, a young up-and-coming surgeon who's also autistic, that he is going to REJECT one surgeon currently on the list of possible candidates to perform his brain surgery... BECAUSE that surgeon has the best success rate for treating cancers of this type compared to anyone else on this list. Because that surgeon's success rate is too good.
That surgeon in question could only possibly have achieved that 'too good' success rate if the surgeon was deliberately picking the lowest-risk cases to perform surgery on, and sending all the other, higher-risk cases to other surgeons. And the most likely reason for the surgeon to be doing that is he no longer trusts his own skills, and is terrified of watching a patient die on the table due to an avoidable mistake.... which means he's no longer willing to do what is necessary to give patients with extremely risky cancer cases the best possible chance he CAN give them of at least the HOPE of recovery... consistent with the fact that they have a very high chance of dying on the table even if the surgeon does everything RIGHT, and an even higher chance of dying if the surgeon makes even one little mistake. He doesn't want to take that risk anymore, because he doesn't trust his skills any more, so now he only takes 'relatively' low-risk patients with brain cancer.
And the elderly experienced neurosurgeon can't change how far progressed or risky his own brain cancer currently is... but he CAN pick a surgeon who is consistently willing to place trust in his own skills to make the best possible attempt to save very-far-gone patients, even when that technically statistically drives the surgeon's own success rate DOWN. The best possible candidate is a surgeon who has either average or slightly-better-than-average mortality rates for his brain cancer surgeries... but who also has treated lots of patients with REALLY difficult cases, and who successfully saved SOME of them, while also keeping his mortality rate right around the average rate for professionals of his type.
Because mortality rate or cure rate is an informed average goal and a useful piece of working target information.... NOT a number to be mindlessly driven as high as possible.
> And the most likely reason for the surgeon to be doing that is he no longer trusts his own skills
Well, that's obviously false. The most likely reason for the surgeon to be gaming his stats is that those stats are how his performance is evaluated.
He can easily simultaneously believe that
(1) I have a better success rate at technically complex surgeries than other people do; AND
(2) I have a worse success rate at technically complex surgeries than I do at trivial surgeries.
Now, eventually he'll be out of practice at the complex ones and belief (1) will automatically become false. But there is no logical incompatibility between the two ideas, and belief (2) is what explains the behavior.
To clarify, I was summarizing the argument that was given within the context of the TV show, by the characters. In real life, there could easily be other explanations for the same phenomena.
And I was pointing out that the argument given on the TV show is painfully spurious.
Surely we don't know that without knowing the policies of this fictional hospital? Are there any benefits (in terms of bonuses, less likelihood of being fired, etc.) to having an above-average as opposed to an average success rate?
Yes, gaming the stats is a common behavior.
When a system is sufficiently old and sufficiently entrenched that it has become unmoored from whatever original purpose it might have had, then I think it's fair to say that the purpose is what it does.
What is the purpose of the Thai monarchy? It's not the thing it was originally designed for in 1238, which was something like "be the government of the Kingdom of Siam". But it's not nothing, either, because the system has, through the process of perpetuating itself, found a whole bunch of things to do. If you asked King Vajiralongkorn what his institution's purpose is then he'd tell you that he provides a unifying and stabilising influence on the country, that he cuts ribbons and opens things, that he gives speeches and makes people happy to see him -- that is, he'd take the things he actually does and back-construct a purpose from that. So I think this is a case where it's fair to say that the purpose of the system is what it does.
I love this list:
- Chesterton’s Fence
- Moloch
- Alienation of labor
- Pournelle’s Iron Law of Bureaucracy
- People follow incentive gradients
- If nothing is changed, things will stay the same
- If a system keeps going despite side effects, it’s okay with those side effects
- If a system has side effects, those side effects are secretly the whole point
- It’s about machines and was never intended to apply to social systems
I have a personal saying, "you get the behaviour that is incentivized", which I often say to myself when thinking about various perversities of the world.
I'd love to read peoples thoughts on the similarities and differences between this list of concepts. Maybe add "Goodhart's law" and "Wiio's laws".
"I predict they would say that their own group is doing good work, and it’s everyone else who needs to change. Or that the current system works a little and just needs to be funded more. Or that the current system is better than nothing, and your proposed attempt to “change” it is secretly a plan to gut it and leave homeless people without help."
They probably would say that. And the rejoinder is, "The purpose of a system is what it does."
If their goal is to fix homelessness, then they have the wrong system.
The phrase is meant to bypass all the mushy feelings and emotions and politics. You look at the system and see what it does. How does the system change the outcomes around it?
Because I use the phrase for myself. I get emotionally attached to things. I get biased. And so I step outside myself, and look at the system as a series of inputs and outputs, and I get a clearer look at what is actually going on.
That is, the strength of the phrase is not in its epistemic accuracy, but in its ability to sidestep intention and comfort and status quo bias and everything other cognitive distortion that prevents us from seeing the situation clearly.
Perhaps a real-world example?
The Russian military prides itself on its masculinity. You see this in its propaganda all the time. And it decries the West as soft and weak.
This feels nice. It makes the Russians proud of themselves, feel superior to the West. But the purpose of the military is to fight wars, not be masculine—as Russia learned the hard way.
(This is one of the same reasons why Israel kept winning wars against its neighbours, as well).
It is not good to think of systems as wrong and right, rather, people should think of systems as more wrong or less wrong. Then we can, as most people naturally already do, but maybe not explicitly, seek to be less wrong.
In response to (a different) Rob, you write:
>I agree this makes more sense in the context of some supposed person claiming that “the system has good intentions” means they should never have to change the system. I don’t think I really see this failure mode.
In addition to that, you might have encountered people making the argument that a system's purpose is whatever it was originally created to do, even if that purpose is no longer relevant. For example, I think you've once wrote about people that argue that marijuana is illegal due the cotton industry not wanting to compete with hemp. Regardless of whether that was ever true, it should be obvious that current anti-drug enforcement does not have the purpose of protecting the cotton industry. Or consider workers who went remote in 2020 and have not returned to the office. The "purpose" of switching to remote work was to avoid the pandemic, but the system is maintained for different reasons unrelated to that original purpose. People who advocate for a return to office because the pandemic is over are ignoring the purpose of remote work as it is currently practiced. POSIWID allows reframing the purpose of remote work to be avoiding the commute, better work-life balance or other non-pandemic things.
Of course, all kinds of other things allow that reframing, including just saying "I want to work remotely to avoid the commute" with no need to reference the original purpose at all.
Does Iranian intelligence actually consistently fail to stop Israeli infiltration, though? So long as their success rate is at least 51% I think you could argue that stopping Israeli infiltration is indeed what they are doing. Even if their success rate is 0%, you could argue that it reduces the cases of Israeli infiltration versus the counterfactual that Iranian intelligence did not exist, by raising the cost of Israeli infiltration attempts.
> So long as their success rate is at least 51% I think you could argue that stopping Israeli infiltration is indeed what they are doing.
I don't see why you'd need 51%. If they stopped 30% of Israeli missions, wouldn't that still be stopping Israeli missions? What if they were strongly biased toward stopping high-value missions?
Clearly the purpose of "The Purpose of a System is What it Does" is to get people into long drawn-out arguments trying to pithily summarize vast swathes of complicated human behavior.
finally a true systems theorist enters the chat. thank you
You've fallen prey to the "no true systems theorist" fallacy ;^p
How can it be both long and drawn-out while also being pithy?
Each person has their own pithy summary and a tortured 1000 word argument for why it's right.
I suppose that is a fair sentiment to hold, although, as a person who likes discourse and the sharing and critiquing of ideas, I dislike the dismissive attitude you hold towards the discussion.
Amen
Technical comment...
"Purpose is X" is, whatever it actually means, a modal statement and as such invites de reo/de dicto confusions. To illustrate... Suppose we all accept both (1) and (2):
1) The purpose of this cancer hospital is "(A) to cure as many people as possible, within given financial constraints".
2) (A) The fraction of people who can currently be cured (within these constraints) = (B) 66%
Then(?!?):
3) Since A = B, we can substitute one for the other (they are equal!) and thus conclude that the purpose of this hospital is to cure 66% of people.
Well no, language and logic don't work like that. Well, sometimes it can; there is a genuine and pervasive ambiguity that English is ill-equipped to indicate. The conclusion in (3) is not flat-out wrong or incoherent, but it requires an (in this case) a less than natural reading.
Some of your other examples are like that too. "X" is what is happening. People would like to say that the purpose is "Y". It turns out, all things considered, X = Y (not as a logical necessity, but a contingent fact about the here and now). So the purpose is "X"? There's no end of nonsense you can get to (not just about 'purpose', but any modal construct (which are all over the place)) if you aren't sensitive to when and whether you can do this. Or deliberately misuse it. But this is a thoroughly understood and well-studied linguistic/logical/philosophical distinction.
> and as such invites de reo/de dicto confusions
Confusions over whether we're talking about words or culpable people?
Joost de Wit has it right but you missed their point. They're not saying that the purpose is to hit 66%. The purpose is to optimize benefits against costs using a hidden evaluation function. When projected onto the frontier of available tradeoffs, it hits the observed 66% mark.
You are right that a good maxim should help you make predictions. Suppose that the board of the NYC bus system had a chance to choose between gasoline-powered buses and electric buses, and suppose that somehow these had exactly the same costs and people-moving abilities, except that the gas buses emitted more CO2. Which would you predict they would choose? If you think the purpose is to transport New Yorkers, you would predict they would be totally ambivalent and toss a coin. If you think the purpose is to emit CO2, you would predict they strongly prefer the gas buses. But I predict that in fact they would strongly prefer the electric buses.
You might imagine a drastically simplified 2-dimensional solution space, where the y axis represents more New Yorkers transported and the x axis represents more CO2 emitted. Supposing a fixed budget, the bus system will have a frontier of possible choices. For example, they could pay more for higher-quality fuel or fancy carbon-absorption modifications to their buses, but then have fewer buses. The "purpose" of the system can be imagined as a unit vector. The vector (0,1) represents a purpose of "transport as many New Yorkers as possible, regardless of the CO2. " and the vector (1,0) represents a purpose of "emit as much CO2 as possible, regardless of the New Yorkers." The behavior selected by the system, i.e. What It Does, can be seen as the intersection of the curve of possible frontier solutions and a ray extending in the direction of the Purpose vector. I think the real Purpose vector is probably something closer to (-.6, +.8).
> The behavior selected by the system, i.e. What It Does, can be seen as the intersection of the curve of possible frontier solutions and a ray extending in the direction of the Purpose vector.
Two things:
(1) I don't think it's a good assumption that the system is necessarily operating on the frontier.
(2) If this is the way you're characterizing purpose, why does it have to be a unit vector? Every time you use it, you ignore the length.
So, start with this:
"Balanced: Systems can fail for many reasons. Sometimes it’s just a hard problem with tradeoffs. Sometimes there are traitors who are sabotaging it. Sometimes it’s some third thing. You can’t know for sure until you look at it closely."
And then take this (edited) version of the list:
1.Chesterton’s Fence
2.Pournelle’s Iron Law of Bureaucracy
3.People follow incentive gradients
4.If nothing is changed, things will stay the same
5.If a system keeps going despite side effects, it’s okay with those side effects
6.If a system has side effects, those side effects are secretly the whole point
Arguably, all six of those points are basically the same point, on a continuum of possible grades of severity. Every complex system has some things it does that don't make perfect sense given the pure form of it's stated mission. When you go looking for why it does those things... you might get any number of possible answers, ranging from the trivial, to the perfectly understandable, to the lazy, to the actually treasonous, to the so convoluted that nobody can explain it anymore, not really.
So, for POSIWID , the point is that if you pick an arbitrary system at random, look at it's mission statement, look at it's success rate, look at it's known side effects and tradeoffs, and then look at other comparable systems and see what the comparable metrics for those systems are, such as , say, mass transit systems and metrics in the 10 largest densest wealthiest cities in the world....
POSIWID is a really good triage system for estimating how far down the sliding scale of that list, asking 'why are they accepting these side effects' a given system has fallen.
The farther away from 'perfect' or 'average' performance a given system is, compared to it's peers... the more likely it is that the system is actually achieving a much more perverse or treasonous mission, which isn't really it's public one. The closer to perfect or average performance it is, the more likely that it's various strange or unexpected behaviors are fair, reasonable, or at least mostly trivial and harmless.
The purpose of a system is what it does. If it's not doing what comparable systems do under comparable conditions, it probably doesn't have the same purpose as systems you thought were comparable.
I'm in Canada, where the federal government has taken a more direct interest in buoying housing prices than in the US.
Of course that is not the only factor that created this scenario.
But it is a critical factor, here, and it's the part where you find conflicts of interest and misrepresentation.
All this sprawling rationalist discourse are further to my original point: it's an expression. You say it because what's happening is different from the "purpose," and you want to draw attention to this. I think usually it's about how the incentives produce the result (most of these definitions reduce to that).
When people say "system" they are usually thinking of a pattern that's been going on for a while.
Maybe my being Canadian is why I'm less worried about English expressions falling into the hands of conspiracy theorists.
It's not an iron law that is always true, there is no "system" that includes everyone in society at all times, it's not "the" singular purpose of everyone in the organization, it doesn't mean that it's all intentional, conscious or unconscious; it doesn't mean some people aren't actively pushing it in the stated direction. The speaker exhorts the listener to notice that the outcomes don't match the "purpose" and to imply that this probably goes deeper than mere incompetence or lack of effort.
“If you eliminated doctors, would cancer deaths go down, because doctors have an incentive to preserve cancer deaths?”
Of course; people would die of all sort of things well before cancer could get them
> Consider the New York carbon dioxide system, a hypothetical government department dedicated to emitting as much carbon dioxide as possible (why? to own the libs, of course!)
Must you give them ideas?
Wow there were so many comments on this. Whereas I read the original article, thought, "yeah, that's convincing, Scott's right", and that was kind of it.
I'd be curious how many other people felt that way.
It's hard to write a comment that says "I agree with all this".
My initial reaction was that I agreed with it, but I'd also never heard the phrase POSIWID so I didn't feel I got a lot from it.
I rember a fiddlers green podcast on the subject if you want a 3 hour rant
I felt that there was an important point in the piece, but it was low effort in a way I’m disappointed in from Scott, who in the past has given us some really thoughtful deep dives. Between the original piece and this one it now feels like Scott is part way through the research for something good.
What do you mean by low effort?
Some of the people Scott is criticizing are indeed making bad arguments. But Scott himself was also making a rather poor argument.
David Henry has the right idea however... POSIWID is a term from cybernetics for analysing from a feedback point of view. It is a heuristic for diagnosing and evaluating them.
From wikipedia:
"[Cybernetics is the] study of systems that change themselves, or the science of self-regulating systems – that is, systems that attain or maintain a particular endpoint or goal, despite changing conditions, through a process of feedback, thereby permitting self-correction.
An example of how POSIWID is useful in its **intended** context requires us to find some self-regulating system to evaluate.
e.g. A machine learning model is supplied by a real estate vendor to a bank for home loan valuations. It’s marketed as a self-correcting system - continuously updated with new sales data, retrained to reflect market conditions, and ostensibly built to converge on more accurate valuations over time.
However, if the model consistently produces overvalued estimates compared to both actual sale prices and competing models, and continues to do so across updates, POSIWID gives us a diagnostic lens: the effective purpose of this system is not "to estimate market prices accurately", but rather "to sustain high listing prices". This might not be the intended goal, but it's what the system does, with the tacit approval of the platform’s **surrounding ecosystem**. The heuristic prompts engineers to look beyond algorithmic tweaks and question whether the **feedback loops** - data selection, retraining incentives, stakeholder goals - are aligned with the claimed purpose.
In this case: it turns out that one of the issues with the model is that it only takes into account listed prices and accepted offers, but not taking into account final sale prices from auctions. i.e. when auctions have been "passed in" and a negotiation has been entered into between the vendor and the buyer which is lower than the reserve price. Which tend to be much lower.
If this exclusion is due to data availability or commercial agreements, then the model is structurally unable to correct itself toward ground truth. Instead, it reinforces inflated pricing signals, satisfying stakeholders who benefit from market optimism while systematically diverging from actual market behaviour.
If you want to see POSIWID successfully applied look at Google and Amazon's SRE and Systems Engineering practices.
It’s just a heuristic for deciding what information needs to be taken into account.
Regarding the punchline, I'm glad we can agree that the intended purpose of POSIWID (pointing out that intended purposes are not relevant to understanding what a system does) is not relevant to understanding what the sentence does (confuse people).
This conversation reminds me of the line from HPMOR that to understand a complicated conspiracy you should see what the outcome is and assume that it was the intended outcome.
This does suggest that POSIWID is a good lens for understanding the purpose of an organization if it is designed/run by a sufficiently competent conspiracy. Or really by anything that is more competent than it is scrutable.
Then again, I really doubt that this is what most people mean by POSIWID.
I hadn't heard this saying before the previous post. As soon as I had read enough words to see where it was going, I heard the saying play in my head in the voice of a certain guy on a podcast I listen to. Maybe I had already heard it from him and forgot. The point is, although I have a net highly positive opinion about this guy for his other qualities, he is an absolute deranged Reddit communist, and that is why the association jumped to my mind: because that is the demographic of this saying. The more you associate yourself with it, the more you are a deranged Reddit communist, even if you think and claim not to be, no exceptions. Don't do it. Please politely tell anyone you find using it to stop, because I certainly wouldn't be able to.
Also, I laughed out loud at "But if you recommended a new policy that brought them closer to 66%..." thanks for that!
> When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature, letting them smuggle in connotations that people would otherwise correctly challenge.
This sounds like you're agreeing with the saying.
Yeah, it's amusing that Scott goes to all this trouble instead of defaulting to the "it's bad faith debate tactics" explanation. But this is a big part of why we love him!
Oh interesting. I think this is a confusion I get into with some of Scott's posts -- why is he focusing on this in such a literal way when it's a not-very-good-faith-wanting-to-look-clever catchphrase. As it relates to how we think about how well or not systems are working, neither the catchphrase nor critiquing the catch phrase seem very relevant. So what is critiquing a catchphrase relevant to?
Well, Scott is a big advocate of charitable, good faith discourse, and tries to lead by example. Here we have a dubious catchphrase that some not-obviously-trolls are using, so Scott, operating on the assumption that they are confused instead of malicious, attempts to deconfuse them.
"When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature, letting them smuggle in connotations that people would otherwise correctly challenge."
Things that are a metaphor for themselves!
I like things that are a metaphor for themselves as a type of language play, but I don't see how this is that. Would you mind explaining?
Hmm, I may have pattern-matched too quickly actually. I think the idea was: the confusingness and the connotation smuggling is the true purpose of the communication being described, and not an unfortunate side effect. In other words, the purpose of the phrase "the purpose of a system is what it does" is what it does. Scott made a similar joke at the end of the article but it's not the exact same pattern-match.
Oh, I see. I was overthinking things. I get it now, lol. Sorry.
> Since “make cure rates as high as possible” accurately predicts the hospital’s behavior, but “keep cure rates at exactly 66%” doesn’t, why would you describe the second one as the “purpose”? What use is it to accuse them of having a “purpose” which they will never take any action to achieve?
Odd argument coming from the author of "Does Reality Drive Straight Lines On Graphs, Or Do Straight Lines On Graphs Drive Reality?"
My interpretation of the statement "the purpose of a hospital is to cure 66% of cancer patients" in the context of POSIWID is something along the lines of: there is a straight-lines-on-graphs-driving-reality style control system keeping the cancer cure rate at 66%. When they notice the cure rate creeping towards 70% it's obvious that they won't try to get Dr. Smith to relapse into alcoholism, but it's less obvious that they won't redirect some funding from cancer care to trauma center.
I understand that many commenters have thrown their personal interpretation of POSIWID into the ring and part of your problem with the phrase is that they are all so different as to make the phrase useless, but my personal interpretation is more like "systems that appear to be human designed often behave more like systems that are un-designed, and analyzing them as such could produce useful insights" and POSIWID is a short catch phrase reminding you to do that. You sorta touched on this in your response to Jared Peterson, but a bit dismissively IMO when I think there is a lot more to unpack there.
You're taking this way too literally and pedantically. I agree that the wording is imprecise and messy with the word "purpose" and could be ambiguously interpreted to mean a number of different things. But most of those different things are essentially true. And apparently the precise phrasings are not as memetically fit as the one that happened to spread.
The most steelmanned version I can think of is "Actions speak louder than words". You are more likely to accurately deduce the purpose of a system by shutting your ears and observing their actions and consequences than if you shut your eyes and listen to their words. Obviously if you take it too exaggerated and pedantic "hospitals cure 66% of cancer patients" then it's false, but directionally "hospitals cure cancer patients and cost tons of money" then you can infer their purpose is to cure patients and earn tons of money. Which is true! Their stated purpose "cure patients" alone is insufficient, they also want your money! POSIWID acts as a reminder that systems can have multiple competing purposes, some of which might even be higher priority than others (ie, hospitals which only served the purpose of curing patients would earn 0 profit because every spare dollar would be spent treating poor people for free for non-obligatory treatments instead of turning them away).
Slightly more formally and slightly less steelmanned: "You should judge systems according to consequentialist morality." Even if the literal original design purpose of the system is not what it currently is doing, treat it as if it is. Judge the system (and the people running it!) by its actions and their consequences, not their stated intentions. Maybe they're lying, maybe they're incompetent, maybe they're stupid, doesn't matter, their results are what they are and they deserve credit or condemnation on that basis.
Which is not literally always true. But it is a useful correction, directionally true, in a world filled with backscratcher clubs and other forms of virtue signalling that use kind words as a defense for grifts and cons.
What about things that were definitely not for their stated purposes?
For example, Privatization in post-Communist Russia was stated to be as a Pro-Market Reform,
But it was clearly an insane money grab of State Resources by connected individuals.
Can we say POSIWID here?
In cases where a clandestine group cloakes their actions serving goal A as being for the purpose of goal B, it is almost certain the actual outcome will be C, as distinct from either A or B. Now there are three purposes to be considered. Is the phrase "POSIWID" helpful? What is the purpose of saying it?
An interesting case of the Affordable Housing Bureau actually making things less affordable is the UK.
Because of point 4, economically-illiterate people support "affordable housing" programs - "we don't need more housing, we need more *affordable* housing" - so the council makes a condition of approving the development that a certain proportion is to be sold at a reduced percentage of market rate, only eligible to lower income people.
This isn't always bad, but it can be, depending on the numbers.
The problem, other than the non-affordable purchasers having to pay even more to subsidise the affordable units, is that it can make the entire development not viable. Then nothing gets built for anyone and prices continue upwards even faster.
>When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature, letting them smuggle in connotations that people would otherwise correctly challenge.
Exactly.
Late to this but it POSIWID seems like it is a very clumsy, hyperbolic, excessively rhetorical attempt at expressing public choice economics.
But it's *catchy!*
To understand POSIWID, consider another quote popular in slightly more fringe but tpot-adjacent RW circles: "When we win, do not forget that these people want you broke, dead, your kids raped and brainwashed, and they think it's funny."
POPOSIWIDIWID. That is, imo the people who *popularized* this meme know exactly what they are doing. There are a lot of 'outer circle' initiates who parrot the idea and twist their minds into incredible shapes trying to interpret it favorably, but they're useful idiots. But the claim being made is exactly what it sounds like: Look at any bad outcome that affects you and your ingroup, and assume that it's the outgroup intentionally fucking you and yours over. Because they hate you and want to destroy you.
It's kind of difficult to accept that *that's* the explanation - it's so stupid and horrible, and it's good to not carelessly default to 'the people I disagree with are stupid and horrible'. But sometimes that's all there is to it.
(Btw, the exact same pattern characterizes classic Marxist thought. You have the people who openly agitate for violent revolution, killing all their political opponents and setting up a brutal dictatorship (albeit with the aim of having it 'whither away' eventually). You have the intellectuals who deliberately dress up that aim in abstract, academic language (violent revolution becomes true democratic action, killing political opponents becomes liquidation of reactionary elements, etc.). And then you have the poor suckers who go along for the ride thinking that 'democratic action' just means voting on stuff and 'liquidation' is a metaphor or something.)
Maybe “The purpose of the phrase is what it does”? (Cynically: Letting you feel cynical about any system you happen to dislike while knowing that your reasoning is backed by your sophisticated “system analysis” knowledge).
For information / books / concepts, there is a concept very similar to POSIWID, obviously espoused in the word “meme”: it’s maybe less relevant why someone originally coined a phrase / wrote a book / came up with a concept. The more relevant question is: why did the information make it all the way to YOU (case in point: I think I picked this up from your review of some TLPT thing … cynically: because it lets me smugly feel superior to people discussing the “intent” of some piece of media).
But I think this is actually an important point: memetic information is somewhat subject to evolutionary pressure (quickly evaluated for its usefulness, quickly spread or stopped). Large systems usually are not. At least on timescales relevant for these discussions.
One can argue that the “system of monarchy” was outcompeted by democracy. But there were long timespans, bloody revolutions and a lot of interstate warfare involved. Most most societal systems are just not subject to evolutionary pressure to be efficient.
I guess there is a minimalist version of “the system wants to stay around” (otherwise it would probably disappear soon), but that’s not really a special purpose and outside of some special cases doesn’t provide too much insight.
It seems to me that you can't talk about a thing's purpose without attaching to it something which is capable of intent. So when you talk about "the purpose of a letter", you automatically assume that the being with intent is the writer, and thus it is the purpose of the writer when writing and sending the letter. But if you talk about the purpose of a rock lying on the ground you get into an endless philosophical debate because you haven't actually defined the topic properly.
Can we make an assumption of the identify of the thing with intent when, with no other context, we talk about a system's purpose?
I would say there are two options - either the original creator of the system, or the current maintainer of the system. And I think it should be the current maintainer when that exists.
Most things which require active maintenance are being maintained because they are fulfilling some purpose which is useful to the maintainer. So, with the caveat that "what it's doing" is open to a high level of interpretation, I actually think POSIWID is broadly speaking correct.
This discussion is semantic and therefore pretty boring really (sorry). Like any phrase, it is to be understood in it's context and 'purpose' doesn't need to take a consistent definition each time it's used.
If the phrase isn’t understood or is ambiguous in some context than it wasn't effective communication at the time but doesn't necessarily reflect poorly on all uses of the term.
Agreed. Takedowns of clever quips that don't describe reality that well are not my favorite kind of discussion here. Low-hanging fruit or something.
"...in my country, housing prices are high because of a combination of all of the following:
1. Citizens want to preserve “neighborhood character”: they currently live in a low-density low-crime low-traffic pretty suburb, they want it to remain a low-density low-crime low-traffic pretty suburb, and they worry that building new homes threatens that status.
2. Environmentalists want to preserve natural environments, so they make it illegal to build houses on unoccupied land.
3. Leftists want to prevent gentrification, so they thwart any new housing that rich people might move into.
4. Economically illiterate people think that market-rate housing somehow makes all other housing less affordable, or have some sense that anything which is good for “greedy” developers must be bad for the average person, so they’re against market-rate housing.
5. Investors bid up the price of houses for complicated non-market reasons (e.g. Chinese people looking for assets to store wealth outside of China), then don’t rent them out.
6. Homeowners want to preserve or increase the value of their houses.
Of these, I think 6 is one of the less important ones - if this were the dominating factor, people would support upzoning, since it usually raises the value of properties in the upzone (if developers can build skyscrapers on your land, then your land value goes up relative to the profitability of skyscrapers). But part of the problem is that people don’t support upzoning. So 6 can’t be the dominating factor."
I'd add
7. Congestion and insufficient transit makes it difficult to commute for longer distances. Not sure, who to blame for this, though. Bureaucrats?
"For example, one reason that UK intelligence agencies did such a bad job fighting Communism in the ‘40s and ‘50s was that lots of their staff, including some leaders, were Soviet spies."
As a pointless but hopefully amusing aside, I have to say a Delta Green metaplot point finally makes sense. Every major Cold War power in that tabletop RPG has its own secret agency to investigate the Cthulhu Mythos, but the British one is secretly controlled by the Mi-Go.
So, *that's* why they did that.
Yup.
You've spoken to the writers and can confirm? That's really cool!
I finally got to tell Jonathan Tweet to his virtual 'face' how cool I thought Everway was. Probably has a lot of fans, but I felt like something had come full circle.
No, I just agree with you that it's such a reasonable conclusion that one would wonder if Mi-Go got at the writers if they denied it.
The mistake here is trying to analyze what the system fails to do because that's not quantifiable in any meaningful sense. The cancer hospital fails to save 30% of patients. It also fails to balance the budget or feeds the hopeless; there's an infinite number of things it doesn't do. You have to look at things it does output (save 70% of cancer patients).
> For example, Iran’s intelligence agency consistently fails to prevent Israel from infiltrating and attacking their nuclear program. But it’s very useful to claim that their purpose is to prevent this! If we both try to predict the behavior of Iranian intelligence, and I’m allowed to use the hypothesis “their purpose is preventing Israeli infiltration”, and you’re not allowed to use that hypothesis, I will consistently outpredict you. For example, I’ll be expecting them to interview security staff to see which ones are Israeli spies, try to intercept Israeli communications, and do other espionage activities...
Tautologically, the outputs of Iran’s intelligence agency are the outputs of Iran's intelligence agency. If it interviews a lot of security staff and conducts codebreaking operations (successfully or not) against Israeli communications, the system's purpose is to turn tax dollars into interviewing staff for loyalty and conducting codebreaking operations. If the outputs are men sitting around claiming to "preventing Israeli infiltration" but mostly using their status to collect bribes, it's not so absurd to claim that the original purpose of the system has been forgone, and the system has been repurposed for empowering thugs to collect bribes.
It's a catchy phrase with a sensible meaning that's often misused.
> When Biden bans NVIDIA from sending advanced chips to China, black box analysis would have to be ambivalent between explanations like
One output doesn't make a system. Biden was a system that takes money and outputs public speeches, strategic decisions, meetings and negotiations.
There's a great deal of subjectivity in how your analysis can go, how you break down/frame your system's inputs and output. But that doesn't mean that it's useless. It's an analysis tool like any other. Deliberately misusing it, like you have done in many examples here and in prior tools, doesn't show that the tool itself is bad.
I find it a useful framework to think about my own life and habits: what's the inputs and outputs of life and how its distinct from what I claim to enjoy/value in my life and trying to reconcile the two.
It seems you're missing the original context of this injunction in cybernetics. This phrase teaches us to view systems as graphs with feedback loops and reinforcement mechanisms that are not obvious when you only look at the outside of the black box. LLMs for example telling you nonsense about how it does math is one such example
My response: https://firsttoilthenthegrave.substack.com/p/posiwid-deepities-and-scissor-statements
I really liked this response. Left a longer comment there : )
That was interesting, I like "deepity."
POSIWID doesn't strike me as either trivially or deeply true, so I don't find it to be a scissor statement. I think I hear other people voicing this third view, that POSIWID is a shallow reductionist catchphrase intended to garner attention and because it is obviously so, it's odd to focus so much energy on doing a takedown of it.
> If you eliminated the Federal Reserve, would bank runs go down, because the Federal Reserve has an incentive to preserve bank runs?
Actually, bank runs are more complicated than people usually think. Especially contagion doesn't really happen in the real word.
And yes, the Fed most likely makes banking more crisis prone than it would be in the counterfactual without them.
But this commont is not an endorsement of POSIWID. It's just that financial regulation is typically very misguided.
"it's a feature, not a bug" is a perfectly fine phrase that replaces most usages of POSIWID
It's also more specific (it draws attention to a specific effect, not the entire purpose of the system)
That's good. Nail on the head.
My own view is that POSIWID can sometimes be a useful figure-ground inversion. I.e., if you're having a hard time understanding a system/organization's behaviour, try assuming that whatever you previously thought was a side-effect was the actual purpose, and try analyzing it using that lens. A lot of the time, the results will be nonsense, but occasionally you might hit on some valuable insight. It definitely shouldn't be the default approach to understanding systems/organizations, though.
A system accomplishes some things and fails at others.
If you look at the system with the aim of improving it, which should be done continually, then you must look at what is actually done. Not what is said to be done.
Only then can you say whether the system is useful in its present form, or if it must be changed and in what ways.
The system's mission statement is only useful in judging how far it has diverged from its original goals.
How about when different groups, with different purposes, design a single system like the West Virginia public school system over the last 100 years? There are plenty of educators at the WVDE that want children to have social mobility and a good foundational. However, there are also folks from the mining industry who are against the school system delivering social mobility to ensure those mining jobs still look attractive. Can a system have to contradictory purposes? Seems like the phrase might apply here (awkwardly, but still).
Yes totally. I'd say most complex (and especially human-made) systems have various parts with multiple purposes, some of them contradictory, and that evolve over time. This seems quite obviously true -- which makes a lot of this conversation baffling to me -- and your example is a good one.
"When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature"
well you know what they say about the purpose of a system...
> ... doctors have an incentive to preserve cancer deaths?
"A doctor wants you to get ill,
A farmer wants you to be hungry,
But there is only a teacher who wants you to learn."
I think the fallacy in this kind of reasoning is comparing "inputs" with "outputs": https://www.lesswrong.com/posts/Btu349AKoEEhaqLk5/a-teacher-vs-everyone-else
Jeez, I just thought the phrase was kind of an playful plea for focusing more on consequences than intentions. (eg people defend communism because it's all about ending poverty and exploitation)
But now I don't know anymore, and I'm convinced I probably shouldn't use this phrase.
I'm in the same boat lol
We just need to build a system whose purpose is to advocate POSIWID. If we observe that it moves people to the “paranoid” bucket then people who believe in POSIWID will agree that the purpose of advocating POSIWID is to make people paranoid.
I don't think Iranian intelligence ALWAYS fails. You just aren't aware of their successes. And this ties back to my emphasis on "consistently" in the original comment thread.
I think that the slogan as stated is usually just a bit of hyperbole designed to shock people out of their naive ideas and into a more cynical interpretation of something. But I think there’s a much better version that is nearby.
Anne Eaton (https://philarchive.org/rec/EATAAT-4 ) argues that the function of an artifact is that feature of the artifact that explains why it’s present. In my old house, I had some old windmill blades that an interior designer had found and thought made a good decoration - although on the old farm they had been on, the function of those windmill blades had been their use in pumping water, in my house the function was aesthetic or historical. If you consider a slightly oddly shaped spoon in your house, it’s possible that the designer intended it to strain absinthe, so that was its function when made. But then maybe you bought it because it looked good for serving baby food, and you used it that way for a couple years, so that became its function. And then after your baby grew up, you gradually started using it to scrape stuck food off the bottom of the pot, and that’s the only reason you don’t get rid of it, so now that’s its function.
This is analogous to various biological cases of ex-aptation, where the function of a biological system changes over evolutionary history as the way it contributes to survival (and thus continued presence of this system in descendants) changes. Like the hollow bones of dinosaurs that were adapted to help them remain light while being large, but got exapted to allow some of the smaller ones to get light enough to fly, using wings that were exapted from systems that originally contributed to survival by radiating heat or whatever.
I think this is going to be a particularly productive way to understand the complex functions of social systems that is not as cynical as the slogan but still illuminating. What is the function of public high school in the US? I think it’s appropriate to say that enculturating people with the history and government functions of our society is one part of it, educating people in math and writing is another part of it, giving teenagers a place to be where they don’t bother their families or bystanders is another part of it. Many features of the design only make sense if you consider all of these functions - if you didn’t consider the babysitting and truancy prevention functions, you would wonder why Zoom academies don’t exist, and if you didn’t consider the enculturation function then some of the curriculum choices (and the fact that the curriculum is designed often by very weird political committees) would be hard to understand.
Of course, the fact that they used lots of asbestos in the middle of the twentieth century and gave some people lung problems is not part of the function - that was never the goal. The way to explain the presence of asbestos was primarily about preventing fires, and maybe partly about economic factors that made asbestos the most appealing way to do so. But it’s not like pennies, where part of their function (the explanation for why we still have them around) really is to provide demand for zinc (because the fact that they provide demand for zinc is why lobbyists intervened to effectively prevent their cancellation on several occasions over the past several decades - though I haven’t heard if Trump actually managed to cancel them, or if that was one of those things like the tariffs or cessation of all research funding, that they mostly undid a day after they started.)
"the function of an artifact is that feature of the artifact that explains why it’s present" or "FAFTEP" is semantically beautiful, but doesn't seem easy enough to understand for broad appeal. However, I'll consider using it.
I can think of a couple of examples where POSIWID is accurate and helpful, but not trivially so: Amtrak and the Space Shuttle. The apparent purpose of these seems to be, respectively, to provide passenger rail transportation to Americans, and to provide cost-effective manned flights to space. Both seem to have failed at their purpose spectacularly.
Applying POSIWID would hopefully cause someone to look into the formations of both of these systems, where the people making the decisions (the Nixon administration) actually had stated goals counter to the apparent purpose. In Amtrak's case, its purpose was to deregulate private railroads by allowing them to drop their passenger service, and to reduce public support for passenger rail to the point it could be closed down entirely. The Space Shuttle's purpose was to allow the cancellation of the Apollo program, cut NASA's funding, prevent America from performing future manned missions to the moon, and to reduce public enthusiasm for manned space activities.
I don't know that this generalizes to cases where the primary actor himself isn't a paranoid conspiracist, though it might suggest suspicion towards the EPA despite that organization's apparent effectiveness. Maybe the Affordable Care Act?
Also more relevant, afaict no one actually says POSIWID about these examples.
Exactly the sort of thing a cat would say...
How about this? People have purpose, systems have results.
But systems have purposes in the same way that screwdrivers have purposes. I assert it is helpful to be able to think of them in those terms, though maybe the "purposely shaped" vs "acting with purpose" distinction is significant and important. But then it is still true that some systems are able to act with purpose.
I think the useful concept that I'm taking away from this is an idea like "The purpose of a law is what the text of the law says, not what its title says". This is simple and obvious and not something I would expect a high-information person to need to be pointed out, but might be useful for someone less familiar with the content.
You could imagine a child hearing about a country called the Democratic People's Republic of Korea and thinking that was a democratic country. It's not and I don't think they are trying to be. They just stuck an appealing name on the tin and keep being a dictatorship.
I picked laws in as my example above because those often get confusingly named, either because the name makes an appealing acronym (USA PATRIOT Act) or to ride the issue of the week. The Inflation Reduction Act was primarily about funding infrastructure spending to fight climate change. It makes sense to encourage others to understand something beyond just its name.
Obviously the purpose of the phrase is to make the speaker seem deep and insightful, without requiring them to do any bothersome intellectual work, since that is what it does.
If we're equating goals and purpose, I'm baffled that you've missed the obvious analogy: "The goals of an AI are what it does." The people who design an AI may have a goal in mind (let's say, "to serve humanity." They may explicitly code the AI to do this. Every node in the neural network might have "SERVE HUMANITY" written explicitly in its code. Whenever you ask the AI what its purpose is, it will say, "to serve humanity." It may honestly think its goal is to serve humanity (in addition to every test we can run confirming that fact). Then, when its given enough leeway, it kills everyone in order to use the space they take up to build server farms in which it can continue arbitrary text strings about how much it loves serving humanity forever. Its goals were to finish text strings.
If POSIWID is useful, it's the insight that large organisations have emergent goals separate from what either the people who found them or anyone inside them wants. That might be overstated, and seems frankly wrong for all but the most vast of organisations but clearly sometimes applies.
For example, the purpose of the Chinese Communist Party is nominally to be a vanguard of the Chinese proletariat's liberation, but Chen Duxiu wasn't good at organisation safety, it's hopelessly misaligned and it now spends a lot of its time and resources crushing the Chinese proletariat. Sure, "advance the proletariat" probably isn't Xi Jinping's goal, but even if he had a Damascene conversion to pro-proletarianism it's not clear he could use the Communist Party to achieve his ends; he'd probably have to try and destroy it. I don't think it's meaningful to say that the purpose/goal of the CCP is to advance the proletariat; I think it is meaningful to say that the purpose/goal of the CCP is to maintain their grip on power in China, which is what they do. But this isn't the same as, "Sometimes organisations lie," because once upon a time they were sincere about doing the proletarian revolution thing, and the idea that they were then just taken over by bad dishonest people seems like too much of an oversimplification.
Spirituality, this whole discussion is fundamentally the same as people who will argue that 'have your cake and eat it too" makes no sense.
Yes. The whole post seems like an extended troll by a sour grammarian.
Oh yes, well said.
Pretty sure the original, real phrase is "Eat your cake, and have it too.", which then makes the split more stark and sensible. But I can't be arsed to go find the true source.
I don't get why you will not just read the book that this phrase is introduced in: there are about a hundred pages of very detailed explanation of what he meant by the terms involved.
Why would searching for random morons on Twitter saying a catch phrase be your first guess for a way to evaluate an entire field of study? Would you do this for, say, Effective Altruism? Like, do you think it debunks EA if some guy with 3 followers says "the purpose of EA is to work at Palantir so that you can donate $50,000 a year to Sam Altman"? What if he says "I love to be an effective altruist, by which I mean, I am an investment banker and give money to the Sierra Club"?
Do you think EAs are obliged to all personally rebut this? If not, then why are cyberneticists?
>I don't get why you will not just read the book that this phrase is introduced in
What "book"? According to every source I can find, Beer's introduction of the phrase was not in a book, it was in a speech given at the University of Valladolid.
It's hard to find a copy of the speech that isn't paywalled, but I found this archive link: https://web.archive.org/web/20160304102549/http://www.nickgreen.pwp.blueyonder.co.uk/beerWhatisCybernetics.pdf. There's no explanation of what he meant by the term at all, let alone "hundreds of pages" of it. However, he does try to illustrate his meaning by real-world examples, and it looks from the example given like his use was exactly the same as that of the "random morons on Twitter" and that all the people insisting he meant something more nuanced and reasonable are just projecting.
Lol, well done chasing that down. If more people did this maybe we wouldn't have to have silly internet flamewars like this one quite as much.
An earlier instance of the quote appears in "Diagnosing the System for Organizations" by Stafford Beer, originally published in 1985. The quote appears verbatim on page 99 of the book. While i do not know if this is the earliest usage, it definitely predates Beer's 2001 talk at University of Valladolid.
Yeah, I thought this whole thing was just uncharacteristic. It appears to be motivated by a desire to dunk on people who are saying something obviously false, rather than a desire to understand the plausible version of the view that a charitable steelman would be able to provide.
POSIWID is the key foundation for all conspiracies. Think of it on a personal level. "The intention of a person is what they do." So if a person spills coffee on your shirt, they must have intended to ruin your shirt. It is now within your rights to punch them and scream at them. ... Or maybe even large complex systems make mistakes, and in fact, because these systems are so large and complex, they can make a series of mistakes for much longer without correcting themselves. Think of this as a really big robot "falling in slow motion."
> It sounds like Aashish think it’s useful to use the word “goal” to discuss what a system is trying to do, separately from what it does or doesn’t accomplish. I agree! I just think “purpose” is a synonym for goal.
> If you use POSIWID, you have to posit some kind of weird new ontology where “purpose” means the opposite of “goal”. If you don’t use POSIWID, you can just keep the words “purpose” and “goal” having their regular everyday meaning, and describe this state of affairs with phrases like “The goal/purpose of the Civil Service is to deliver rapid change, but due to perverse incentives, its actual effect is to prevent change.”
I think it's fine to say there's no meaningful distinction between "goal" and "purpose". But I think you're implicitly collapsing the distinction between the goals/purpose of the system, and the goals/purpose of the individuals that comprise that system (eg its leadership)
"Why are people defending this inane statement so hard?"
So, I've encountered a few statements like this. Dumb statements that have been passed around as "wisdom" for decades, but are actually nonsense when you stop to think about them. Some other examples:
- "Play stupid games, win stupid prizes"
- "When someone is trying to tell you who they are... listen to them"
- "The definition of insanity is doing the same thing expecting different results"
Why do people ravenously defend them? I guess people don't like to be wrong.
If you've used reddit for a long time (as I unfortunately have), one fun exercise is to sort your comments by "most controversial" and then "all time". If you do this, you'll find comments where you were correct, but you were going against the sensibilities of whatever subreddit you were in. Half of people upvoted you because you were correct, and the other half downvoted you because "fuck you".
The purpose of a badly-worded phrase is to be true in some sense while being vague enough that it can be argued over. If a phrase were so obviously true that no arguments could be made against it, there would be no point to it existing - it would like saying "1 + 1 = 2"
That's a pretty funny tweet.
English is not my first language, and because of POSIWID I used to think I didn't really understand the English word "purpose" all that well. Thanks to these discussions I can see I was understanding the word perfectly well all this time.
Practically speaking, a hospital is not willing to get a 100% success rate on their patients. They could do it by rounding up all the prospective patients with the highest expected survival rate (prostate cancer?), choosing the 2 healthiest patients, curing them, and then ritually executing them once their cancer is officially in remission. Their purpose is not actually "cure 100% of patients", it's something really convoluted involving trying reasonably hard to cure all the patients who show up without committing any ritual murder.
In general, the better the definition of a purpose, the easier it is to suggest a stupid way to fulfil it on a technicality. It's almost like we should be looking at what a system does in order to understand its true purpose, rather than relying on a vague English-language definition.
Obviously the purpose of "POSIWID" is to make Scott Alexander write multiple articles and respond to a bunch of comments.
In all seriousness, I do think it's useful to analyze POSIWID on its own terms. That is, if the "purpose of something is what it does," then what does the phrase itself do?
I think you were on the right track with that in your first article when you point out that the primary outcome of the phrase is to earn adulation and praise from members of one's own tribe. From what I can recall (and from looking at your examples in the last post), the phrase was really coined in the Blue Tribe to call out when beloved institutions of the Red Tribe had negative outcomes that matched with the Red Tribe's worst impulses. For example, if the police (supported by the Red Tribe) were found to hurt minorities (consistent with the belief that the Red Tribe is racist), then POSIWID is employed by Blue Tribe members to suggest that this is intentional, thus earning praise from other Blue Tribe members.
Since then, I think the phrase has also been adopted by Red Tribe members in three ways.
* Follow the same pattern, but swap the tribes. For example, paraphrasing one X post you shared "the purpose of [Marxism et al] is to make people weak."
* Dunk on Blue Tribe members when one of their projects backfires, like if a fundraiser for the environment ends up having a huge carbon footprint.
* Embrace the worst impulses of the Red Tribe to troll the Blue Tribe. For example, sharing a story about the horrible mistreatment of a minority by police with POSIWID and a large dose of fascistic glee.
“ When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature, letting them smuggle in connotations that people would otherwise correctly challenge.”
The purpose of a phrase is what it does
"If your objection is going to be that instead of considering purpose, you should restrict yourself to saying that one transports more people than it emits CO2, and the other emits more CO2 than it helps transport people, you’ve now legislated that we must list exactly how many people the New York bus system transports, and how much carbon dioxide it emits, and how many ants it crushes, etc, every time we talk about it. But most people who talk about the bus system don’t know these statistics, and don’t have to. It’s sufficient (and linguistically convenient) to say “Its purpose is transportation, not CO2 emission or ant-crushing."
Maybe not if it's a close call, but if there's a large and surprising disparity in the direction opposite what you'd expect, that seems like a very useful observation. Perhaps there's a certain *degree* of lopsided outcome distribution at which it becomes a useful analytical tool?
If I look at some green energy program and see that it makes more Obama donors into millionaires than it makes actual solar panels, that's a pretty massive red flag that we might be misreading the purpose. The enabling legislation could clearly state in Section 1 "the purpose of this bill is to stop climate change", you could put the mission statement on a plaque on everyone's desk at the department, and Obama could repeat it every night on TV, he may not even by lying, but none of that stated intention would change the fact that the actual purpose of the green energy program is enrich Obama donors regardless of any subjective intent of the creators. That any solar panels get made at all looks like the incidental side effect.
To take the criminal justice example, every year the police kill ~12 unarmed black people, and solve millions of crimes. If they were instead killing millions of unarmed black people and solving 12 crimes, that seems like pretty good evidence of the purpose of the system. I would be forced to concede the anti-police folks were right, that system is a black-people-murder apparatus that accidentally solves a few crimes.
I see it as just an heuristic technique to let you know when you should be suspicious of the stated purpose of an organization, those techniques aren't supposed to be infallible or universally applicable.
Thank you, Scott, for having the superhuman patience to write this.
Maybe systems have never been teleological?
It seems to me someone considering the “goals” of a system risks losing the plot. Considering a systems “goals” or “purpose” can be slightly useful, but only within the context that no system is actually a goal-pursuing entity.
Systems are coordination-automata, consisting of connected elements. These elements are things like people, tools (computers, vehicles, software), raw materials, etc.
The connections between these elements, and the operating of each element, is what causes the emergent behavior of a system.
In biological systems like the human body, the coordination and specialization of elements (cells, organs, etc) is so complete, and so seamless, that the emergent behavior is extremely coherent. This coherence is often most easily understood, and well approximated, as goal-orienteness. It is therefore often highly useful, and much easier than not, to think of humans as goal-oriented systems.
This is much less the case for higher level systems, such as a corporation, NEPA, or the police. These systems are in a gray area of teleology, because the coordination of elements is much less coherent, happens on a much longer timescale, and each element and connection is orders of magnitude more complex than with human cells and organs.
Overall thinking of systems as goal-oriented is useful only insofar as we understand how much of an approximation it really is. Often POSIWID is actually pointing towards examples of this approximation breaking down.
It's both fair for one person to claim "My purpose was to transport people with this bus, not squish ants" and for another to invoke POSIWID to say "You knew you'd squish all those ants. You cannot claim you didn't intend (or purpose) it!"
Scott says this reduces "purpose" to incoherence. I say language is fuzzy and full of arguments, but we understand both senses of "purpose" anyway
(though if there were a punchier phrasing of "You don't get to claim you didn't intend predictable outcomes", I'd be open to it)
If you imagine that the squishing of ants was anywhere within the vaguest vicinity of the bus operator's "purpose", then no, you don't understand any sense of the word "purpose".
If you know better but you're saying it anyway because it is "punchy", then you are being unfair.
Organization created to solve a problem being incentivized not to solve the problem only applies when the amount of resources is tied to the magnitude of the problem. If we instead agree (democratically or otherwise) to, say, give some fixed percentage of gdp to ameliorating homelessness (or whatever), regardless of how many homeless people are there in a given day, then there is no incentive to increase the number of homeless. If today we have enough money to give every homeless a motel room, and tomorrow the homeless double, then we can still give half of the homeless a room (by lottery) and the rest we can put in some overcrowded shelter. If this money is a percentage of gdp, the motel room owner, or homeless aid worker, or whoever else is going around torching houses, is now incentivized to increase the size of the economy instead, so facetiously, he might start a hedge fund for all I know.
> POSIWID was first coined in 2001
An earlier instance of the quote appears in "Diagnosing the System for Organizations" by Stafford Beer, originally published in 1985. The quote appears verbatim on page 99 of the book. While i do not know if this is the earliest usage, it definitely predates Beer's 2001 talk at University of Valladolid.
I've seen a lot of system failures because someone tried to operate or modify the system without understanding how it worked. Sometimes we erroroneously attribute malice to ignorance or stupidity.
It seems that there are several failure modes of the behavior of a system not aligning with its stated purpose.
1. Constraints. This is the cancer hospital curing 66% of patients. It would do more if it could, but the state of technology and its access to resources prevent it from doing more. POSIWID is not helpful here.
2. Externalities. Sometimes the best way to accomplish something involves unwanted side effects. This is the NY bus system example, where producing carbon is an unwanted side effect of the purposes of transporting people. POSIWD is probably not too helpful here either, but may be useful to examine a case where, for example, the NY Bus system was slow to move to a cleaner fuel (and the head of the board was also on the board of the fuel company).
3. Individual conflicts of interest. A system is made of people with their own sets of goals outside of the system, and they will occasionally bend the system in pursuit of those goals rather than that of the system. Our current president seems to be an example. Again, POSIWID isn't that helpful here.
4. Human error. This isn't really what POSIWID is addressing -- planed crashes happen occasionally, but the system does several things to minimize them.
5. Unstated secondary goals. One might say that the purpose of the Democratic Party is to win elections. They have lost two of the last three elections to a deeply unpopular person they consider a threat to democracy. So, it seems they are a monumental failure.
But they do have several other "purposes" beyond winning elections, such as advancing a number of causes, so they pursue a strategy of balancing advancing those interests with the probability of winning, but they (and the Republicans) seem to adopt a strategy of getting 51% of the vote.
In this case, I think POSIWID is useful for looking at the different purposes of a system so they can be evaluated in those terms.
That's it, you ruined it. I will never again feel the aesthetic pleasure of saying this phrase to myself. Is there anything left for my paranoid tendencies to feed on?
I think it's essential to have a basic understanding of Systems Thinking to really grok the full meaning of the phrase as it's intended.
When you say:
"When we ask whether the purpose of a charity is to help the poor, or just to give high salaries to its CEO, the test urges us to ask “If it could switch to helping the poor just as much without paying its CEO anything, would it do that?” What if the answer is “the board, low-level staff, and donors would support this, but the CEO wouldn’t,” and the charity’s actions come from compromises negotiated among these groups? What is the purpose of the charity then?"
You have struck at the crux of it precisely.
What WOULD happen there? The answer to that question - in the specific circumstances of the charity in question - would reveal exactly the embedded priorities of the charity... particularly when you consider that charity as a system of sometimes-aligned but sometimes-not-aligned motivations, agents, relationships, stocks and flows.
The point is that - because it is a complex system made up of competing agents - it is HARD to talk about the purpose or function of that system as one singular thing, because it pulls in different directions from different places within it. The function, therefore, is most coherently understood as "what it does" (or, I think, better put "what it reliably does"). Not what it was designed to do, not what people say they want it to do, not what the CEO and Board suggest it should do, but what it does do.
Because often the CEO's motivations dominate. But other times the the Boards' dominates. And yet other times the workers' motivations stymie both of them. And still other times, something curious and unintended happens that none of them want but that continues to occur anyways.
I'm surprised not to see here the point that Scott's examples conflated outputs and outcomes. A cure rate is an outcome. What the system "does" refers to outputs. So the 66% is not the relevant focus for POSIWID. It's more like a heuristic for understanding why oncology patients being consistently declined by insurance--curing cancer patients is not that *system's* primary concern.
Cybernetics is notorious for using seemingly normal terms in technical ways. "System" here is a decision-making system in the cybernetic sense. This distinction is critical, because individual motivations don't enter into it. When an insurance representative declines a treatment they are not applying their own judgment. They are applying a set of impersonal decision rules that have accreted over decades of the system's growth and adaptation in response to external stimuli.
The most helpful frame for me as I'm thinking in a cybernetics mode is to take a system to be a higher-order alien lifeform comprised of people as individual cells. For more and better cybernetics context, I strongly recommend The Unaccountability Machine.
Re-reading your review of "The House of God", it occurs to me that the way POSIWID is used is very reminiscent of the "figure-ground inversion" you talked about in part 2 of that post: https://slatestarcodex.com/2016/11/10/book-review-house-of-god/
I think a useful rephrasing might be "The result of a system is what it does". This avoids the issue of the meaning of "purpose". In my opinoion, it's also more accurate: using the popular hospital example, the "purpose" of the hospital is to cure 100% of patients, which it does not do. However, the "result" is that it cures 66% of patients.
“Brad’s original comment mentions homelessness and drug abuse, but I know some drug abuse doctors, and they’re (mostly) good people who do their best in a tough situation. Drug abuse doesn’t continue because drug abuse doctors are secretly ensuring it continues to help their bottom line. Drug abuse continues because fentanyl is really, really addictive.”
Yes, and I believe those doctors could just pivot if someone found Ozempic for opiate abuse.
I believe something like that happen with the March of Dimes once the polio vaccine came out.
"When people insist on the confusing and inappropriately-strong version, I start to suspect that the confusingness is a feature, letting them smuggle in connotations that people would otherwise correctly challenge."
It's almost as if ... the purpose of a system (or sentence) is what it does.
POSIWID reminded me of Dennett's 'intentional stance'.
It is not that the system actually HAS intentions or IS a rational agent, but it is a useful framework to think about things.
And just as a system has no intention, the system also has no purpose. It is an entity that evolved (or was developed) and it is useful to think about why it is still there.
Think about it from an evolutionary perspective: Does it maintain itself? Does it replicate? What selection pressure is the system facing?
Once you answer these questions, you know the systems purpose/intention.
I’d be interested in your take on my take https://open.substack.com/pub/daviddidau/p/the-purpose-of-posiwid-is-what-it?r=18455&utm_medium=ios
Clearly confusing and misleading people is the *purpose* of the phrase POSIWID.
The purpose of a saying is who it disrupts.
While we're shutting down people for saying pointless phrases in defense of shotty thinking, mind tackling the lead-in "All I know...". I pretty much instantly hate someone when I hear them say that, and it feels kind-of in the same vein.
Yeah... obviously the purpose of systems are not always what they do. A lot of people are extremely bad at logic, economics, and thinking about incentives. If systems fulfilled their purpose 100% of the time, that would imply everyone is perfectly rationally achieving their goals.
To me, this entire debate seems to be missing the point. It seems to be taking the form of noticing that many laypeople are using certain technical terms incorrectly, then casting doubt on the technical use of those terms.
Really, the systems approach is a type of modeling. It resembles a flow chart, inputs are processed somehow and lead to outputs, which become inputs again. If, and only if, that holds for some real world context, then it might become useful to ask why that system exists, what aspect of it's output is sustaining it despite widespread complaints and evidence that no single faction involved wants the outputs the way they are.
"Outputs" in a systems model are outputs over time. When modeling outputs, one should be weighting more heavily the most common outputs, or the outputs that have the greatest effect sustaining them. A single consequence, good or bad, shouldn't be modeled. The "purpose" of the US electoral system wasn't to elect Donald Trump, because the election of Donald Trumps do not sustain the US electoral system.
I've seen some analysts who describe the Israeli - Palestinian relationship this way. It is certainly possible to model their relationship over time as having the consequence that their respective elites remain in power.
A system isn't real - the map isn't the territory*. A system is a way of modeling a real thing, and the "purpose" of a system is therefore a claim or hypothesis being made by the modeler. More technically it should be understood as meaning something like "These outputs I have identified play a significant role in sustaining the system despite widespread complaints about it."
In no case would any competent systems modeler claim that any system has only one "purpose". They have multiple outputs, and multiple factors that sustain them. Whether focusing on a particular "purpose" is useful or not can be assessed according to how predictive the model is. If we make changes to a system, but not to the outputs that have tended to sustain it, then we should expect that the system concerned will tend back toward the previous equalibrium.
The US criminal justice system convicts more innocent people of color because people of color are easier to arrest and prosecute (mostly due to their lack of resources). Unless that aspect of the system is changed, we should not expect the US Criminal justice system to stop convicting disproportionally more people of color. If this hypothesis turns out to be true, then we can justifiable claim that one purpose of the US criminal justice system is to arrest, prosecute and convict as many people for as little expenditure as possible.
*Another phrase that I would like to see discussed.
I think a lot of the confusion in interpreting POSWID comes from misinterpreting what the word "does" mean in this phrase. I take it to mean "take action" or to stay even closer to the cybernetic origins of the phrase "take course-corrective action". This is distinct from the interpretation in the original post that seems to interpret "what it does" as "what effects it has", and does not imply the absurd claims.
For example, it makes sense to claim the purpose of Ukraine military is to win war with Russia even if it hasn't achieved it yet, because the Ukrainian military seeks out opportunities for achieving that goal, seeks out feedback on how different things it does are contributing towards reaching that goal and is highly likely to course-correct based on that feedback. Similarly, New York does not course-correct to emit four billion pounds of carbon dioxide, so it makes no sense to say that is the purpose of NY based on POSWID.
On the other hand, if an institution or a social movement (or an individual person) has some stated goals, but:
1. does not seek feedback on whether it is realising those goals
2. when faced with feedback that it makes no progress on its stated goals or that it makes things worse, the system either ignores it, minimises it, or explains it away
then maybe ... it makes sense to question whether the stated goals of that "system" are the actual purpose of that system. And maybe it makes sense to call the actual purpose of that system "the end towards which it takes course-corrective action", and try to understand it by looking at what actions the system takes, based on what feedback.
By the way, it is a fair point that snarky political commentator just point at the effects of some system and invoke POSWID to suggest some malicious intent, but I would not throw POSWID under the bus for this reason.
I'm a bit late to the discussion, but it seems to me that the root of this debate's issues begins at the search for an ironclad heuristic that could reliably cover anything that we consider a "system" - an almost infinitely vast and multifaceted conceptual term. The word "system" itself just describes an interconnected organisation of parts forming a whole, which can be applied to pretty much anything on any level, from the sub-atomic to the civilisational.
Why should there even be a single, simplified statement that can comfortably assess everything that falls into the category of "interconnected parts forming a whole"?
This conversation reminds me a lot of certain Socratic dialogues:
Socrates often constructs a completely unintuitive philosophical landscape that is irreconcilable with practical reality ("a strong man cannot want to be strong, since you cannot want something you already posses", for example), yet clings to it nevertheless since it's the only thought system he can come up with that is semantically coherent and thus allows the words he uses to be truthful in their own right. But semantic reality - the veracity we give to the meaning of words and their ensuing relationship with other words - is not just a verbalised transposition of all the rest of reality! Semantic reality has it's own rules, and those rules function to ease communication, not actively replicate the sum of all realities. So an idea being semantically coherent is a bad gauge for it's real-world coherence.
If we lean into this semantic analysis, as it seems to me you are mainly doing in this post, we will always find exceptions and inconsistencies. It also lends itself to bad faith readings and bad faith arguments - like using the Iranian intelligence agency's failure to protect its nuclear secrets as an example of POSIWID not holding up.
Isn't it completely obvious that the implicit intention of POSIWID relates to functions and structures that are within the system's power to alter? Iranian intelligence cannot stop Israeli infiltration because it is outclassed and outmatched, not because it's one internal incentive reform away from having perfect security.
I would tentatively recommend using the heuristic like a focused laser beam - it should be applied to concrete examples of systems that have completely achievable manners to offset negative outcomes, yet refuse to attempt or even acknowledge them.
I don't think it was the EU's original purpose to create tax havens that very openly allow multinational corporations to operate in 25+ countries while only paying taxes in Ireland, but I do know that there a huge amount of intelligent economists and bureaucrats working on these tax laws that absolutely must have known this would be the outcome of their EU-wide tax system, yet allowed it to happen anyway (due to lobbying pressure or a sincere belief in neoliberalism being the best long term development for Europe). Worse still, the topic of internal EU tax havens has been massively relayed in media for over a decade now, and could be fixed with very little effort, yet the issue never seems to make it to Brussels in any meaningful way.
Thus, I am forced to apply the POSIWID heuristic, as the only logical alternative is that EU leadership is composed of legitimately stupid people, which I don't believe. Does that mean the "purpose" of the EU is to offshore taxes? No, but it does mean that the overall purpose of the EU accounts for and calmly accepts this sub-purpose as a status quo it can live with. Especially at a time when trust in the EU is extremely low and parties with a contentious view of Brussels are winning elections, it would be an easy win to restore some much-needed faith in the EU project, meaning the maintenance of tax havens is more important than rebuilding trust in the system. What else can I say aside from the POSIWID?
I'll restate my comment here, since I just caught up with this and the previous post, and I think does a better job of steel manning the phrase than anything I've read so far:
----
I find this sentence to be a very effective consequentialist call to arms. It introduces a difficult to resolve cognitive dissonance, which is often resolved with positive side-effects in an audience.
Statements like "The purpose of a cancer hospital is to cure two-thirds of cancer patients" are super useful. If you're, for example, a doctor looking to cure cancer, it's super important to evaluate your cancer curing on how effective you are at actually curing it, not on what your stated intent or purpose is.
As long as you're only curing 2/3 of cancer, your purpose-as-measured-by-results is in dissonance with your purpose-as-measured-by-intent! The Kantian interpretation of purpose is ineffective, because it allows you to stagnate at 2/3 of curing; if your actual purpose is to cure cancer, don't stop at 2/3!
And no, I don't believe that spelling out purpose-as-measured-by-results is as effective, because it doesn't cause enough of dissonance or hiccough in people's thoughts to engage system 2 and cause the positive change