Is this a similar effect to people developing tolerance to low oxygen/low air pressure conditions? Or maybe actually exactly the same effect (i.e., increased red blood cells in people who are in low oxygen environments for long periods of time?)
Is the CO2 in my room really high (possible, even with windows wide open; thanks, wildfires!), or does that word not actually exist in the grid? There is no neighboring "C" and "H" pair on this board.
Extremely cool. One possible way to salvage the original result is if there is a cumulative threshold effect: maybe you need to be in a low-carbon environment for a long time to see effects and if you regularly breathe fresh air your mental function is not affected. This wouldn't explain the submarine data point though
Unfortunately, some of the pro-CO2 papers claim acute effects. Like Satish's experiment used 1 hour of exposure pre-testing (https://ehp.niehs.nih.gov/doi/10.1289/ehp.1104789), and is one of the most-cited ones and has p-values out the wazoo.
I'm not sure it would be that obvious. Recent analysis of court records have shown that judges levelled more severe sentences against guilty offenders the lower their blood sugar, ie. that hungrier they got/closer to lunch, they became hangry. After their lunch their sentencing severity popped back up nearly where they were at the start of the day.
I suppose we'd have to survey lawyers if this was common knowledge, but my understanding is that it wasn't.
It's an interesting example of a claim that spread through headline repetition, when anything other than headline repetition would have instantly discredited it. The effect size found in the study was a *one hundred percent reduction* in the likelihood of being acquitted as you get closer to lunch.
Without any knowledge of lawyers, I'd say they were aware that the guy scheduled immediately before lunch isn't going to be acquitted, and isn't going to take very long - but that they would have provided the same explanation as all subsequent discussion of the paper has, that the court schedule is not determined at random, and that guy was placed before lunch because he was going to be convicted in a short, predictable amount of time.
I found something else more recent which found the effect size to be overestimated* though because of the varying length of cases, instead of time of scheduling, so it may still be debunked. That's as far as I'm reading for now, but there's a lot to it. Thanks for sharing that, I hadn't thought to question it before.
Also, it's a long way from "affects our mood" to "affects our mood in this particular way, under this particular circumstance, all other things being equal, x% of the time."
I recently (June) started monitoring AQ in my condo after reading similar studies, watching a presentation by DHH from Basecamp (MRqh8oLY7Ik) and trying to find ways to help my wife's migraines. I'm using the Awair AQM8002A.
Our overnight CO2 in our bedroom (shared with 3 children) went up as high as 1300ppm and rarely went under 900ppm any time during the day. I tried opening windows and I could push it down to 700 but as soon as I closed the window (since it's July in Houston) it shot right back up. The humidity however took much longer to recover.
We left for a month and it went down to the 400s (set the temperature to 78 instead of 71).
I just used Condensate Pan Treatment Tablets (Uric acid) two days ago and our overnight reading have gone down to high 700s (from 1300ppm). I've had a lot of trouble figuring out how such a large change in CO2 could have been caused by micro-organisms in the condensate pan... My first thought is faulty meter, which I haven't tested.
No way to know the effect on migraines as it's a long term occasional problem. My personal waking fatigue is much improved though.
Is urban dictionary a valid source of words for the game? Narc…
Could there be factors other than CO2 that correspond to ventilation? Air flow, humidity, level of VOC, sound levels, etc.
I sleep much better with a fan. A study might show that people sleep better with fans running. Headline, “Studies show high air flow helps sleep.” Later we find out it’s the white noise that’s helping.
You should go decide which environmental factors to record for each game in advance so that you don't need to play 800 new games every time someone brings up a random factor that is potentially correlated to intelligence.
Some factors that might be worth recording, in addition to co2 and humidity, in no particular order:
- How much you slept
- VOCs
- Ambient sound level
- Air flow
- Stress (as in how much other stuff you have to do soon)
- Sugar intake
- Lighting
- Screen time
- Commute length
- Obesity?
- Chewing gum
- Watching stupid entertainment shows
- Fluoride
- Meetings
- Secondhand smoke
- Whether you took Ambien/Xanax/etc
Also, random thought: If you don't want to play the word game yourself a bunch of times in order to test hypotheses about human cognition, you might want to ask (some of) us, the vast ACX readership, to do it on your behalf.
Recording for each individual game, how many games immediately prior in a session and how you scored, might not be a bad idea. I get into a zone with my certain computer game and my performance usually improves for a few games in a row, then drops and improves again. If you were playing one game a session you’d always be cold, but if you play in blocks this tendency to temporarily improve might obscure whatever the CO2 was doing to your scores in general. Given a repeated task in the presence of some high but more or less constant level, maybe the brain adapts to the situation at the time.
If you are modifying the experimental conditions anyway, and there are no budgetary constraints, I'd suggest following additions:
(1) Air oxygen meter.
(2) Get an app that beeps at regular times (or random times could work too) and asks you to record if you feel like playing or not and record the air metrics. Get automated sensor that record the metrics anyway, so it is less work. Maybe you know someone who can hook the sensors into an Arduino board / RasPi?
The previous data from the same location can also be quite useful.
> I don't think it's likely that it causes the Dutch East India Company.
I also agree with you that humidity is not the ultimate cause of the existence of the Dutch East India Company, but I also think that the phrase "Dutch East India Company" in the above comment is a malapropism.
I think they're making fun of you and BronxZooCobra for using the acronym "VOCs" without ever explaining what it means, since VOC was also the acronym for the Dutch East India Company.
Google suggests "Volatile Organic Compounds". I assume it's something along the lines of pollen count.
You might find this amusing, if you haven't heard of it before: There's an urban legend that sleeping with a fan can kill you. https://en.wikipedia.org/wiki/Fan_death
(I also sleep with a fan. It hardly ever kills me.)
Some spaniard scientist seemed pretty alarmed by the effects of infrasound on airplane crews´ blood vessels at one fascia congress and advised to shun computer fans. But that was at a conference years ago and today I guess there's not much behind it.
My anecdotal evidence is that I feel better and more alert when I ventilate the car on long drives. But in car CO2 levels be much higher than a house or apartment. I can’t think of a good way to test this.
I had wanted to measure this exact same thing - ever since I moved into my current flat I feel that I can work so much better with the windows open. And the price did stop me - I thought these monitors were around $10 when I got the idea (why do they cost so much?). So... if logistics of me being in the UK is not an issue, I'd be happy to test?
Send me an email at scott@slatestarcodex.com with your planned experimental protocol (doesn't have to be too official) and your address and we can talk about it.
If you're actually willing to buy the monitor for someone(s), I suspect you'll find plenty of video game streamers who would be willing to do this as a gimmick on their stream for some weeks, so then you'd just have to decide which game(s) you think are more "cognitition" focuses.
I'm a streamer. https://www.twitch.tv/nebupookins I'm open to playing a wide variety of games on stream (except I don't like to play horror games). I've also bought two CO2 monitors after reading this post (not the one you recommended, a $30 one and a $50 one) with the expectation that if they both report the same reading, they're probably both accurate, but if they report different things, then one (or both) of them are inaccurate.
I'm really curious if you have some suggestions for experimental protocol.
Are you playing any competitive online games? Overwatch, League of Legends, Starcraft? Those usually have pretty detailed statistics for player performance. I play LoL and third party apps like Porofessor show me very detailed stats per match, from "teamfight participation %" to "creeps farmed per minute". But to avoid multiple comparison adjustments, I'd just stick with measuring competitive score progression. If you gain more points when playing in the fresh room than in the stuffy room, this is good evidence for the effect.
I'm streaming too sometimes, a mix of Russian and English streams, gaming and photo editing. https://www.twitch.tv/loweren
I used to play League, but my performance in League already has insanely large variance so not sure it'd be a great metric for me. E.g. creeps per minute varied by a factor of 18.5.
A lot of recent top-level chess has been sponsored by an air quality monitoring company who used this gimmick. Eg you can see it at 1h30m here: https://youtu.be/EmVuLy08Y4k. The commentators sometimes mention it too, in order to help the sponsor.
If any streamers are reading this: I would find this gimmick pretty entertaining, and would direct some money/views toward a streamer who did this in an at-least-moderately-careful way.
How would you make it fun without unblinding the gamer about CO2 levels? Or maybe blinding is asking too much since people can usually sense low oxygen anyway.
People can't sense low O2 - that's why simple asphyxiants like nitrogen and the noble gases can kill. What we sense is high CO2 directly.
(Simple asphyxiants are dangerous because breathing them doesn't give O2 but does remove CO2 - as such, you don't sense "bad air", you just pass out from lack of oxygen and then die.)
Confused though, nitrogen and noble gases aren't converted to CO2 by the body right? Is this part of why CO is so bad? CO would be converted to CO2. Though CO binds to hemoglobin so strongly this effect is probably negligible for CO.
What's the mechanism for this exactly? How do we sense the high CO2 that makes us want to breathe out? CO2 is pretty inert. Is it through CO2 acidifying water? If I put another gas that acidifies water in the air does your body start to process it as CO2, and will you have trouble even if there is enough O2 in the air?
Lots of questions, but would like more explanation here!
I'm speaking about the case where you're breathing pure N2 (or another gas/mixture that while not toxic does not contain O2).
If you're breathing pure N2, you will indeed maintain low levels of CO2 and not be poisoned by it. However, you need a constant supply of O2 in order to live, and pure N2 by definition doesn't have any. As such, after a couple of minutes of breathing pure N2 you will pass out from lack of oxygen (and then die unless rescued).
This form of asphyxiation (no CO2 accumulation, just lack of O2) can only occur if you are breathing gas with low CO2 and also low O2 - this is highly unusual, because "bad air" breathed out by other animals will reach dangerous levels of CO2 long before the amount of O2 drops to insufficiency (40,000 ppm of CO2 is 100x current baseline/150x pre-industrial and immediately dangerous to human health, but that's only 4%; animals use something like 1-1.5x as much O2 as they produce CO2, but even a 6% drop in O2 from the 20% baseline is still 14% - more than half of it is still there). The usual cases for sudden anoxia are sudden decompression of a plane, huffing helium balloons, SCUBA accidents or improper storage of cryogenically-liquified gases - to put it mildly, not the conditions natural selection designed us for.
I'm not 100% sure of the way by which we sense high CO2. I do know that strongly-acidic gases like HCl, NO2 or SO3 are very dangerous to breathe, although this has more to do with direct acid damage to the lungs than with homeostatic errors.
Cool! What's the mechanism by which CO2 is dangerous? If I breathe in something is just like air except with the CO2 in place of the nitrogen, so the oxygen content is still good, what happens exactly? Acidification, some of reaction A + B <-> CO2 + X where the equilibrium is off, something else? Why would I die? Isn't CO2 a relatively inert gas?
It is one of the cases where evolution cuts corners that no good engineer would: it tests something easy (CO₂ → acidity was my guess too) that is strongly correlated *in nature* with the thing it needs to test (O₂).
I learned recently (although I would need confirmation) another example: water. We feel thirsty when we need water, and we get minerals at the same time. We do not have a feeling for the lack of these minerals. Therefore, drinking distilled water will sate our feeling of thirst, but will not supply our body all it needs.
CO is bad because it binds to the heme group in hemeglobin better than O2, so it will displace O2. That is, your red blood cells will absorb CO before they absorb O2. It also induces a conformational change in hemoglobin that inhibits oxygen unloading, so even the O2 that is left isn't delivered efficiently to tissues. CO2 doesn't bind to hemoglobin at all, it's quite a different looking molecule.
Yes, the main sensors controlling breathing in the brainstem sense H+, which increases with CO2 levels because dissolved CO2 largely forms carbonic acid by interaction with water. There are additional sensors that directly detect O2 levels also.
Acidifying blood per se is not thought to trigger increased breathing, because the detectors are actually sensitive to H+ in the cerebrospinal fluid, so something has to cross the blood-brain barrier (which H+ cannot) and *then* cause a decrease in pH. CO2 does that.
Very cool. So, is there any other gas that makes a weak acid and achieves the same effect? I'm thinking about SO2 (which is nontoxic right) for instance. SO2 also gives you the acidification thing? You don't have the A+B <-> X + SO2 thing though.
Haemoglobin does bind CO2, via the N-terminal amino group (which can be carbaminated). It's a minority of the blood's transport capability, and it certainly doesn't involve the haem site, but it's not correct to say Hb doesn't bind CO2.
Don't need to worry about which are more "cognition" focused. Cover a wide array of genres and stream types and see if any have a marked effect. If the effect is indeed specific to a certain kind of decision making, this could tease out what it is.
Personally, I feel like there *might* be an effect for tasks that depend heavily on what people call "emotional intelligence" (scare quotes), but not for anything else (unless you want to talk about long-term carbon dioxide overexposure or something, which is a different topic).
Yeah I'm wondering about coming up with new ideas vs repetitive things. Again this is because a lot of people come up with new ideas while outside walking, and my suspicion is that the mechanism for this is probably not CO2 but it's a reason the CO2 hypothesis sounded plausible to me. I also feel I can focus better when the window is open, but it could just be a placebo effect or it could be that the sound from outside is some sort of white noise, or that having some change in the background stimulates new ideas (maybe it's extra hard to work and come up with new ideas while staring at a white wall) or something else I don't know.
I think there is a lot of interesting stuff in this space that is chronically under-explored, and which needs more empirical data so we can all figure out how to work more efficiently. This seems like a pretty underfunded direction of research...
It would also be interesting to see if there is a "natural" way to split various human activities, and whether differential effects on how well we can do them from various outside influences can be used to classify them. So is there some way in which tasks A,B are more alike to each other than C,D, in that A,B both become harder under circumstance X but not C,D? It would also be interesting to look at the corresponding brain activity while doing all of these but this would be hard unless you had some kind of mobile pocket MRI machine.
Again I think there are a lot of interesting questions here. We should try to understand them better because the upside could be potentially ginormous.
Nice self-replication! And I'm pretty convinced by your successive ruling-out of explanations as to why the effect is real but you failed to detect it (e.g., controlling for temperature).
Sorry if you mentioned this and I missed it, but is there any chance that time (like "index" of the trial/game) was a factor? Like, maybe you got better at the game over each successive play, but CO2 levels were also positively correlated with time? In which case you might mask a "real" effect of CO2 if performance is driven by Time (+) and CO2 (–). As you say, to salvage the effect you'd need for these effects to be of roughly the same size, which is a stretch. But might be worth checking.
Another possibility would be that you're already so good at the game that you're essentially at ceiling, so there's just not enough variation in your performance to explain, even though you observe plenty of variation in CO2 levels.
I don't love this explanation because to take it seriously, you'd need an account where the influence of CO2 is dependent on prior ability in the task. Basically an interaction between Expertise and CO2 on Performance, such that High Expertise is somehow "immune" and Low Expertise is more susceptible to CO2 levels. I guess this is in principle possible, especially if the task is easy enough for people with high verbal reasoning/fluency. But: 1) it doesn't seem that intuitive to me; 2) if it is true it's a pretty big caveat for the initial results, and implies they must've sampled from Low Expertise population to have detected an effect (assuming they didn't regress out Expertise); and 3) now we're just adding epicycles.
If you mean getting better over successive games over the course of months, this is definitely happening, but CO2 isn't changing consistently (other than I guess in the global warming way, which is much slower). If you mean getting better over successive games in the course of a sitting, I have the data that would let me test for that but I haven't done it - just eyeballing it it doesn't look that way, nor does CO2 go up very much over the course of my sittings (I usually play only 3 or 4 games max in a sitting, which takes about 15 minutes, which isn't enough time for room CO2 to increase much)
I'm definitely not at the ceiling - my "score" (which I calculate artificially as score divided by average score for the board) has varied between 54% (my worst game) and 143% (my best) over the past month. The world record for most boards is often around 2x my own score.
Yeah I meant the former ("getting better over the course of months")––I just thought maybe if time was correlated with wildfire season or something, you might detect higher CO2 levels. But that makes sense! Seems like a solid non-replication.
“ Several parallel scenarios are available, allowing retesting individuals without bias due to experience and learning effects.”
This is a quote from the paper that Scott cited. I wondered the same thing as you; whether a lot of prior knowledge of the test affects performance enough to overcome any problems with CO2 levels
The lesswrong post linked about chess is nearly a good measure but is made much noisier by the confounding factor of "how good is the random anonymous person I happen to be playing today?". Happily chess furnishes a much better measure of day-to-day fluctuations in cognition, which would be to solve chess problems: positions (either taken from real games or artificially composed) in which there is a single, objectively correct answer to the question "What is the winning move here"? This also seems more like "decision-making" than finding as many words as you can in a word-search puzzle in a set period of time.
Does chess have a recognized solution to this? I don't know exactly how Elo ratings work, but it seems to involve adjusting for your opponents' skill (eg you get lots of points for beating someone better than you, and you lose very few points for losing to them). Would "average Elo points gained/lost per match" be a fair measure of how well you're doing that day? Do chess websites make that easy to track?
Oh sure, they can calculate and track this automatically, but a whole game of chess consists of so many decisions that you could feasibly lose to one person on one day because you made 40 great moves and then one dumb move (median cognitive strength: good!), and then beat someone else of exactly the same elo strength the next day by making 40 mediocre moves and then one more move that didn't lose (median cognitive strength: average!). The advantage of solving a single problem (ie choosing a single move by calculating a huge branching tree of variations) is that there is no time limit (it can take half an hour or more if you like) and it's a single decision, so it is arguably a better measure of your peak concentrated cognitive power at that time. (There remains inevitably some subjectivity in how to grade the difficulty of problems though.)
Elo is good for this. Given two players' ratings, you can calculate the probability that one will beat the other. So rather than tracking win/loss or +/- points, you can track if you did better than expected or worse than expected and by how much.
Chess.com also provides an accuracy rating (how closely your moves matched the engine's recommendation). This could work too. The only problem is that if your opponent is much worse than you, it's generally really easy to get high accuracy because the best moves are obvious. But with a decent match-making system I wouldn't expect that to be much of a problem.
I disagree with the other person's puzzle suggestion; some puzzles are just harder than others, and unlike with opponents, there isn't really an objective way to account for that.
I forgot to add -- I've played around with chesscom's API and using it to track the rating info would be easy, but unfortunately you have to pay for accuracy info. You could conceivably track stuff like "number of blunders, mistakes, and inaccuracies per move" by hand for free though.
If anyone does decide to use chess as a marker for cognition, I would recommend ignoring win/loss entirely and focusing entirely on moves. Use stockfish or something to compare scores before and after each move. Proportion of good moves or some other move based transform will be more robust to blunders (by you or your opponent) than game win/loss.
Any chance you can test the correlation against different CO2 sensor noise/bias models? I believe you that low readings happen when you'd expect and high readings happen when you'd expect, but I can imagine observing that even if (CO2 reading) = m*(True CO2 level) + b or even something higher-order.
If X is correlated with a linear transformation of Y, then X is correlated with Y to precisely the same degree. If it's not a linear transformation, then the observed correlation could end up being larger or smaller, but to completely erase a correlation that is in fact strong would likely involve a weird transformation.
I did something similar playing Go against the computer when I worked long overnight shifts in an intellectually focus demanding job. Every few hours I'd play a few quick 9x9 games against the computer and decide based on my resulting rating how well I was mentally handling whatever level of fatigue I was in.
Scott: instead of doing univariate p-tests, why not fit a linear model where you add an extra term that depends on days since you started playing to model your skill increasing over time? I’m happy to fit the model for you if you feel comfortable sharing/uploading the data somewhere.
Done! See http://slatestarcodex.com/Stuff/CO2data.xlsx . You'll want to use "ME AS %" as the dependent variable, that's my score as a percent of the average score for each board. I didn't start recording some thing until later so those columns will be blank until I start. Let me know if you have any other questions.
Hi Scott: short answer - there's absolutely no statistical effect that I can see from any of the variables you specified.
Long answer: here is a COLAB that anyone can run with all your data so they can check my steps, I threw this together pretty quick: https://colab.research.google.com/drive/1R8KBD_3Wvh_0KmI3YMcdzkMve6JXeXJ6?usp=sharing - if you are interested in the summary, it's the final cell at the bottom. As you can see, the total R^2 is pathetic (0.13) and none of the covariates are statistically significant.
Oh, and the final model is: 'normalized_score ~ total_days + minute_of_day_sine + C(day_of_week) + CO2 + temp' - with the C(day_of_week) syntax specifying that I used a dummy encoding to capture any effects from day of week.
Really interesting to see this. I started a company to commercialize a CO2 capture material back in 2014. At the time, a lot of people I talked with were excited about this study, but I was always really suspicious of the effect size.
I was planning to kick of some self-experiments on cognition in a few weeks using chess puzzles, working memory, and math tests. I ordered a CO2 monitor and will add that in to the mix.
Anyone have any other interventions or supplements they think would be interesting to test?
The sleep study just finished and I’ll be unblinding the data and analyzing this weekend. That’s what got me thinking about what my next study would be.
I think that CO2 has a big impact on my productivity, because I seem to get distracted much easier and am less willing to do real work when I spend a longer time in a less ventilated room. (In fact, I am likely to play a game like WordTwist in that scenario :D) But I never measured this with a monitor, so I would be interested how well my notion of distracted-ness really correlates with CO2 levels. I just can't think of a good, unbiased, way of measuring it...
No, that game is just very responsive to practice. I started out getting about a third of average, and after playing it a few thousand times (I only measured carbon dioxide for 800, but played it much more), I'm slightly above average most of the time. Most of the people whose plays are going into that average have been playing it hundreds or thousands of times, and learned all the stupid meaningless words that it accepts as real (some of the most profitable words are short extremely obscure ones using common letters, like telt, lari, esne, and tael). You also need to learn the generation algorithm (about half the time, it inserts s a very long word usually ending in -ness or -tion, so you often want to search for those endings and then try to work back from there). With a little bit of addiction and wasting your potential as a human being, you can become as good as anyone else!
I put significant probability on the hypothesis that cognition responds to fast changes in CO2 concentration, but not to absolute CO2 concentration; ie, if you go from outside to a stuffy room, you'll be mildly impaired for some duration, and if you go from a stuffy room to outside you'll maybe get a boost for some duration (or perhaps an opposite-direction effect which is also some sort of impairment), but if you wait awhile in either environment you'll return to baseline. This is what you'd expect given a homeostatic mechanism that isn't instant. This would reconcile the original results (where people go straight into a high-CO2 room, and do worse) and your results (where you let the CO2 concentration rise at a slow, natural rate).
This makes sense. Blood co2 levels are not primarily driven by air co2, and the health effects of low ventilation that building codes are concerned about use co2 as a proxy for accumulation of volatile organic compounds and other gases, not co2 itself. At least climate change isn’t making us co2 stupid!
I will consider testing it, mostly because the game is fun. But also does anyone have any other circumstances to test this with that could affect intelligence? “Before vs after strenuous exercise”, “while fasting vs while not”, “before vs after big meal”
Oooh what about high vs low air pollution days? Although id expect that to fail too, the negative effects of partial combustion. product pollution is probably longer term. Many possibilities
Yeah, and the effects of VOCs aren’t short term cognitive ones. It is in the context of a rebuttal to the claim that co2 bad mask bad being backed by reference to the permissible co2 concentrations in buildings
Within one building, CO2 should be a good proxy for VOCs, but it's possible that either the low CO2 building or the high CO2 building happened to have a really gassy carpet or paint or something.
Not necessarily. Within a single room, local sources and sinks will predominate. For design purposes, or over buildings with tens of thousands of square feet, CO2 <-> VOC is a good (or at least reasonable) proxy.
Did you happen to experience the other negative symptoms associated with high carbon dioxide at the times in which your monitor recorded very high levels?
I agree they're basically the same, but they use different dictionaries, different letter frequencies, and different generation algorithms (Boggle's is random, Wordtwist's isn't) to the point where skill at one doesn't carry over to skill at the other very well, and in some cases you need different strategies.
doubled my scores on day 2, roughly. Mostly by speculatively trying 4 letter combos that sound like they might be a word, and focusing on vowels and suffixes.
I don't think I have a very low credence on substantial individual differences on this. If nothing else, I get the impression that there's a lot of individual variation in how unpleasant it is to be in a high-CO2 environment. Still, for some people to have a large effect and some people to have zero effect would be somewhat surprising.
As a former submariner, I don't recall noticing any effects from varying CO2 levels. I do remember noticing huge effects from varying Oxygen levels as the concentration was adjusted. Lower levels and everyone was tired all the time. Higher levels it was harder to sleep and when I did, I would have bizarre dreams. Maybe there was an effect from CO2 but it wasn't noticeable to the larger effect of oxygen.
In a good way, in that you didn't feel tired all the time, but also this was sometimes a bad thing when all you wanted to do was sleep between your shifts on watch and couldn't "time travel" as it was called. There wasn't a lot to do when you weren't actively on watch, so it could be hard to pass the time.
I think the CO2 was a side effect of crew activity, but O2 was controlled. I don't know this for certain as I was just a worker in the engine room side of things, but it was said they would keep O2 levels lower for periods when we were on mission so that more people would stay in bed and make less noise. The level differences were only noticeable on extended journeys when the recirculation / ventilation at periscope depth was not used as to remain hidden.
Is it consistent with your data that while CO2 has no effect on your performance *given that you decided to play the game*, it may affect whether you're up to the task of doing something vaguely cognitively taxing? In submarines and spaceships, you're in an environment that forces you to do things or else you die/are Severely Talked To by your superiors in an extremely strict social hierarchy; in an experimental room, it doesn't really matter whether you focus or not.
The game is called Boggle, tho that usually refers to the 4x4 version. I was addicted to it in grad school, but frustrated that the boards sometimes had only 3- and 4-letter words, hardly worth the effort of finding. So I wrote a program to generate Boggle boards with more and longer words in them.
First I tried making the probability of a letter showing up proportional to its frequency in English, but that resulted in boards where you could find TENT 3 different ways. Then I tried using a genetic algorithm that scored boards according to the frequency of the trigrams found in them, but that was even worse--it could fill the board with regular tilings of T ,H, and E.
Eventually I had the program count the dictionary words found in each board, and use that as the GA score. That worked well enough to run smack into premature convergence, which turned out to be so disturbingly powerful that it radically changed my views on the impact of technology on society.
(Now I think that rapid communications with high connectivity has taken us into a regime where we're no longer capable of evolving quality cultural artifacts / designs. Modern art was perhaps the first case of cultural premature convergence due to tech--an artistic ideology that spread like fire (their analogy, not mine) along the new rail networks linking Europe, and completely displaced all others in institutional art when it was still half-raw. I think premature convergence also contributed to the decline in innovation in pop music since 1990.)
For your purposes, you'd like to reduce the variance of the max possible score on each board. You could try giving yourself an adjusted score of (your score) / (max score). But I don't trust the scoring system to scale the right way with word length so that this adjusted score would have the same expected value on words of all lengths. Words like "intrapsychically" could mess with your adjusted score. It might be better to write a wrapper around the board-generating program that rejected boards with an unusual distribution of word lengths.
I use max score / average score of other people who have played the board. Usually there are about 20 previous players so there's not too much variance caused by the last player being really bad/good.
Can you explain more of what you mean by "That worked well enough to run smack into premature convergence, which turned out to be so disturbingly powerful that it radically changed my views on the impact of technology on society."?
John Holland's 1975 book /Adaptation in Natural & Artificial Systems/ proved the Schema Theorem of genetic algorithms, which guarantees that, over time, the "schemas" (genetic building blocks) with the best scores will be present in an exponentially increasing number population members. This means that the top-scoring genome is very likely to get fixated in the population very quickly once it has a foothold of just a few organisms.
This would be great if the fitness landscape were pretty smooth and lacked poor local maxima. But if it did, you probably wouldn't be using a genetic algorithm, which is useful largely because it helps organisms escape local maxima (without using randomness just to add noise, like most techniques do--it instead uses its random numbers to recombine partial solutions which each have high scores).
This would (I think) be okay if your genome were so high-dimensional (e.g., human genome, 3x10^9 base pairs) that each individual locus can have only a tiny effect on the score, and even chromosomal crossovers are likely to have small impacts on the score because they contain so many different loci that their scores will have a normal distribution with a very small variance.
But my 4x4 Boggle genomes were just 16-dimensional. So a single mutation could increase one genome's score significantly, and that genome would take over the entire population in a few generations. The population has then converged prematurely (before you ran all the generations you wanted to) on a single (crappy) solution.
I tried different ways of staving off premature convergence:
- increasing the mutation rate
- increasing the population size
- using log(score) instead of raw score
- increasing the mutation rate as the score increase decreased (this is the opposite of simulated annealing!)
Nothing worked until I tried dividing up the population into different groups, and severely limiting interbreeding between groups. For evolution to continue long enough to find good solutions, the population needed safe local niches where new genomes could arise without being instantly stamped out by the one dominant genome.
As I went around thinking about premature convergence, I started noticing it all around me. Starbucks. Tolkien imitators. The rise of a single national Native American accent. Plotless literary short stories about alienated characters who fail to resolve their existential crisis. All 1990s pop music.
Worse still, tech allowed perfect reproductions of books, music, and videos. In the 19th century, music and drama had to be live, and so nobody could capture more than a tiny fraction of the global market. 20 years after the introduction of the gramophone, an article in the March 25, 1908 /Musical Courier/ said that 3,000 professional pianists were starving in Vienna. Today there are perhaps only 5,000 professional pianists, starving or otherwise, in the entire US (https://www.zippia.com/pianist-jobs/demographics). In the 20th century, popularity shifted to a power-law distribution, so the top 20 artists in popular art forms make more money than the bottom 100,000 (I have looked at actual numbers in the past, but may be misremembering by an order of magnitude). Music and acting aren't professions anymore; they're lotteries.
Change and variety have dropped precipitously. English-language pop music changed more from 1964 (Elvis & the Everly Brothers) to 1968 (Pink Floyd & Jimi Hendrix) than it has in the 52 years since (and I'm not even going to qualify that with "might have" or "by some measures"). Literary novels became indistinguishable and boring. So did SF & F. I read My Little Pony fan-fiction now, not because I'm especially fond of My Little Pony, but because I think it's better fiction, and far more diverse in style and perspective.
I think that the pattern seen when new technology suddenly connects previously-isolated communities isn't a gradual slowing-down of cultural evolution. First there's a sudden burst of innovation when everyone is first exposed to everyone else's ideas. That innovation slows down as everyone's experience becomes more and more similar.
A completely different application of premature convergence is in education, where we introduce it deliberately. Most Americans have always, for some reason, thought that ideally every kid in America should have the same education. Evolutionary theory suggests this is a terrible idea, culturally and economically.
Yeah, culture is more complex than Boggle. I think premature convergence is a pattern or tendency, not a Law of History. Also my example of FM radio is sketchy, because shouldn't something similar have happened after the introduction of records in 1889, and after the formation of radio networks in the 1920s? (Maybe it did? I have no idea.)
It seems to me your characterizations of the cultural convergences as *premature* relies on your value judgments. Maybe the fitness landscape in these cases is quite smooth, and the evolution converges rapidly to what is indeed the global optimum.
That's a reasonable hypothesis, but the change in the distribution of popularity in culture (from maintaining population diversity, to being taken over by a power-law distribution of just a few big winners like Starbucks or Beyonce') matches the change in the distribution of genomes (from maintaining population diversity, to being taken over by a few, and then just one, genome) in premature convergence to a lousy global optimum. Culture with high-tech communication and reproduction seems less-able to sustain the population diversity needed for continued evolution.
This is clear in the case of language. The change in English from Beowulf to Chaucer (< 400 years) was huge compared to the change from Chaucer to Elizabethan English (< 200 years), which was huge compared to the change from Elizabethan English to the present (> 400 years). But then language is something that governments WANT to be standardized rather than optimized.
(Also, the Beowulf-to-Chaucer change was mostly driven by a sudden breakdown of linguistic isolation in 1066.)
Statistics on the changes in the distribution of years-to-bankruptcy, over the past 150 years, might be informative: are the big near-monopolies living longer? (My impression is no.) Or maybe a measure of product diversity would be better. But either approach would be plagued by complications due to the (human) population increase over that time, the shift to globalization, the shift to electronic media and social networks, and changes in regulation.
I would like to test the idea, but don't have any good idea right now how to do it.
Hmm. I'm not seeing how you the dynamics of the approach to equilibrium would tell you anything about whether it's a local equilibrium in a rough landscape or the global equilibrium in a smooth. I mean, if there *is* such a clue the protein-folding people urgently want to hear about it ha ha.
I think the only thing that would be solid evidence that it was a local equilibrium thing was if, after some time, the system (pop culture, language, whatever) lurched out of its apparent equilibrium and found some better one -- then we could conclude, post-factor, that the first was a local equilibrium and the fitness landscape was rough.
The inherent problem with this, though, is that we think the fitness landscape is *also* constantly changing, id est there isn't any absolute objective standard for the quality of pop music, it's a question of what the musical tastes of the moment might be. I think we can also make a good argument that the fitness landscape for language also changes all the time (e.g. through population migration, new technology, communication changes).
Using the phrase "solid evidence" here is problematic. There's a lot of territory between "not telling you anything" and "solid evidence". The dynamics of the approach to equilibrium tell me a lot in the case of Boggle board evolution, because I studied hundreds of cases, and the score of the final result was very, very strongly correlated with the number of generations before convergence. Is that proof, as in geometry? Not at all. But it's a whole lot of Bayesian bits of evidence.
The application to culture is speculative, and there are lots of other factors in play which would be difficult to factor out of an analysis. I'd be surprised if it were easy to identify many clear cases of premature convergence in culture. But I'd also be surprised if that dynamic turned out not to be a significant factor in cultural evolution.
An important line of supporting evidence which I haven't yet mentioned is the theory of island biogeography, which gives an equation relating the size of an island to the number of species its ecosystem can support at equilibrium, and a lot of empirical evidence that fits it well. The take-away is that evolution is much more powerful--you evolve many more species per square mile--if the territory is divided into lots of little areas with boundaries between them that greatly constrict gene flow. This seems to be due to the same dynamic I describe as "premature convergence". So it does appear to be a major--probably THE major--factor in evolution, which is a complex, real-world application. This makes it more likely that it's also a major factor in cultural evolution.
You might also want to adjust the score by the frequency of each word in a text corpus. But again, it would be difficult to know /how/ to adjust the score, and the safer solution would be to reject boards with very rare high-point-value words.
My purely anecdotal self-observation is that my cognition powers drop dramatically when I'm tired, whether or not I realize that I'm tired. In fact, noticing that my cognition is impaired is a clue that I must be tired, even though I don't feel tired, so I lie down and am usually out like a light. This must far outweigh any effects of CO2 - unless the CO2 is making me sleepy, which it could be.
For me, the studies in this area are very frustrating to read, because they all track CO2 exposure for different time durations. For example, the Navy did a study (Rodeheffer) of exposure for 45 minutes (presumably because they had some reason for looking at brief rather than longer duration exposure), while the longest time of exposure studied seems to be 8 hours (Allen) (because it's the length of a standard workday, at least supposedly.) In one study (Satish) subjects were exposed 3 times in the same day and measurements done for each exposure (I don't understand why this protocol was chosen). It would be interesting to look at whether there was a pattern to results of first, second and third exposures but if this was in the paper I didn't see it.
Generally it looks like the longer the exposure period, the more likely the study is to see some effect, which might explain some of the differences results between studies that all seem to have been pretty carefully done. With this WordTwist informal experiment, however, we have no information for duration of exposure prior to the test, so I think the results can only be interpreted to negate cognitive effects of brief or momentary exposure to various levels of CO2 and can't be compared to exposure for longer periods of time. I don't know why it's meaningful to test momentary exposure to CO2 at these levels, but then again, I'm not understanding why any of the exposure times were chosen, except choosing 8 hours to replicate a standard workday.
Show your data. Show *all* your data. This would all fit so, so nicely into a single scatter plot, and I'd be about 5x more convinced if I could see it (especially if you can work in time data in a way that lets us know how your learning the game looks compared to other effects).
I’m not a statistician, but isn’t p=0.97 a really high p value? My understanding is that if you have a completely uncorrelated thing your testing, your p value should be a uniformly distributed RV. So there’s like a 5% chance of p<0.05 and same for p>0.95. I guess this isn’t a real critique beyond “if I’m understanding the stats correctly this seems so high it makes me wonder if something’s up”
Ah, so it’s like how you need a high powered study to get a significant result when you’re dealing with a small effect size. Since this study has so many samples, if there actually is no effect it makes sense for the lack of an effect to show up more clearly?
Thanks, I forgot the difference between N=800 and N=30 lol
No, bored-anon is confused and you were correct initially.
P-values are derived by comparing the observed deviation to the calculated distribution of sample means. The predicted distribution of sample means becomes narrower with a larger sample, which causes a given deviation to result in a lower P-value.
The chance of P > 0.97 is indeed 3% regardless of the size of the sample - getting P = 0.97 with N = 800 involves a much smaller deviation than the deviation that would give P = 0.97 with N = 30.
I'm an environmental engineer, but I don't specialize in indoor air quality. It is not a climatic impossibility to have indoor CO2 levels < outside air CO2 levels. A local sink (most likely a houseplant) will easily accomplish that. Whether it's reasonable for your word gaming ecosystem is for you to determine.
Possible. You could test by placing your CO2 detector outside near the plants on a sunny day (and not during a drought - plants should not be water-stressed) and seeing what it reads. Ideally at some different times of day, too. If reading >> what you were seeing inside, it's not those plants causing the low reading. If reading is low vs typical atmospheric, doesn't prove it, but at least indicative.
Back of the envelope, sunlight is 1000W/m^2 and plants have a photosynthetic efficiency of at most 1%, so you're getting at most 10W/m^2 worth of CO2 converted to glucose. 10 joules = (10/4184) kcal = (10/4184/4) grams of glucose = (10/4184/4/180) moles of glucose, which requires (10/4184/4/180)*6 = 0.0000199 moles of CO2. So that many moles of CO2 are being absorbed per second by one square meter of the most efficient plants in direct sunlight. Meanwhile, one cubic meter of air is 1000 liters, and 22.4 liters at STP is one mole of air, or 44.64 moles per cubic meter. So if the plants were sealed in a 1 cubic meter greenhouse the CO2 PPM should drop by 10^6 * 0.0000199 / 44.64 = 0.446 per second. So it seems plausible that if you're in a very green rural area and the wind is blowing the right direction you could get a legitimate CO2 reading significantly below the worldwide average.
But google tells me "2000kcal / 24 hours in watts" is 96, so you need 10 square meters of the most efficient plants in direct sunlight just to keep up with the CO2 produced by one human. It seems very unlikely that this condition holds in any normal indoor area.
Could the plants reduce CO2 below outdoor levels while the house is empty? Suppose a typical house is 200m^2 with 2.5 meter ceilings = 500m^3 and has the equivalent of 1 m^2 of the most efficient plants in direct sunlight. Then the plants would reduce the CO2 PPM of the unoccupied house by 1/500 of the rate in that hypothetical greenhouse above, or about 3PPM per hour. It probably wouldn't get very far below the outdoor numbers before ventilation puts it in equilibrium.
Roughly half of all outdoor readings will be below worldwide average. There are diurnal and seasonal cycles due (at least in part) to plants, as well as human activity. And the human activity causes substantial local variation - this is why Mauna Kea measurements are preferred for tracking long-term global CO2 change. Here's a Chinese paper suggesting daily cycles ~30 ppm and seasonal cycles ~60 ppm in Beijing (note minimum is in summer): https://pubmed.ncbi.nlm.nih.gov/14551950/
Thus, the baseline to throw out data should at least be locally measured CO2 minima, not any global number.
I remember reading that houseplant VOC destruction, previously estimated to be very large, was in fact mismeasured, and more recent research had reduced the magnitude of the effect substantially. Can't find a citation with quick searching, though.
Assuming your math is correct, a houseplant may not be the only way to locally sink CO2. I'd have to think for a bit to say that there's nothing else. Globally, the other big sink is the ocean, but probably not enough water in a room to make a difference.
Yeah, the basic test for whether indoor plants matter is "are your plants adding mass in comparable quantity to the amount of food you eat?" (technically, the amount of [food - faeces] you metabolically process). Of course, plants only consume CO2 during the day and actually produce it at night, but if the mass gain of the plants is orders of magnitude less than how much food you eat then they can safely be ignored.
This is why motile plants don't exist; photosynthesis doesn't provide enough energy for locomotion. Animals get the required energy because they can tap a much larger area of photosynthetic activity than their own body area - a cow sustainably eating 6 hectares of pasture is (inefficiently) harvesting the entire photosynthetic production of that pasture.
How does that CO2 monitor compare with the "smart" one from Awair? Await keeps a record of the CO2 levels over time, so it might be easier to use in a study. I also don't know how accurate it is, but can tell that opening the window and the like affect the CO2 quality.
Andrew Gelman has a great term: “trying to weigh a feather in the pouch of a jumping kangaroo”. Sure there can be a ton of confounding variables, but if this were a meaningful effect (for certain definitions of “meaningful”) it wouldn’t require a research study that takes those into account to assess it.
Secondary problem is even if you did find an effect it doesn't account for adaptation effects. My prediction would be your body optimizes for a given CO2 level over some period. Consider climbers adjusting to higher elevation--takes a day or two.
It might apply more to motivation than calculation.
It's totally plausible that CO2 homeostasis mechanisms include throttling back the production of it, rather than just breathing more to get rid of it and excreting it in urine.
Throttling back production might involve the brain trying to avoid work when blood CO2 levels are elevated. But I don't know if 1000ppm is enough to have a measurable effect on blood CO2 levels.
"Generally, under normal physiologic conditions, the value of PCO2 ranges between 35 to 45 mmHg, or 4.7 to 6.0 kPa"
At atmospheric pressure slightly above sea level, 100kPa, this is equivalent to 47,000-60,000 PPM. So an extra 1000 PPM in the air you're breathing is not likely to matter. It'll slow CO2 clearance by 2%
It's likely other factors that correlate with having windows open are causing the inconsistent effects in some of the studies effects: both lighting and odors can have emotional effects which might affect performance.
Also as an Ashkenazi Jew your ancestors may have spent a 99th percentile amount of time indoors and picked up some adaptation, if there ever was any effect of CO2 per se. In hunter-gatherer and farming times maybe high blood CO2 sort of signaled "I've been exercising too much and need to rest" but after urbanization it meant nothing.
As a practical matter going for a walk outside in the fresh air feels good and elevates my mood. I don't know what the active ingredient is.
When you said high CO2 levels could affect focus and decision making, I thought that made a lot of sense - in the context of continuous high CO2 levels. e.g. if you work in a closed-up office but go out at intervals for 'fresh air', you might feel refreshed and able to continue working, but if you stay indoors all day you might feel much worse.
Your experiment seemed closer to the scenario of sometimes going out for fresh air - I'd be much more curious to see an experiment with one group working in the same high-CO2 environment every day, and another in a better-ventilated one.
Watch out for CO2 meters that "calibrate" their sensor by assuming that the lowest level in any 7 day period is outdoor air. This is a terrible assumption, and gives entirely bogus readings that change over time, but that's how a lot of them work.
Instead, you want to get one with dual IR sensors, one of which is used infrequently in order to calibrate the other. (The theory here is that the main source of decalibration is the IR sensor's LED dimming with repeated use.)
One way the competing factors could cancel despite large effect sizes is if your inclination to play is or is affected by a threshold effect. Like a more complicated and convoluted version of measuring the temperature in a room when the thermostat of a heater kicks in, and then concluding that opening the window doesn't have an effect on the air temperature.
You best hope there is no cognitive problems caused by high ambient CO2 levels. The US Navy has several Boomers on station, underwater, with very high CO2 levels at all times. Any one of them could kill a couple of hundred million people, and trigger Armageddon.
I firmly hope, and fervently pray that our brave sailors manning those vessels are not cognitively impaired.
Indoor air quality is typically significantly worse than outdoor air (fire season aside). Seems like a plausible explanation here is that CO2 is a proxy for how well ventilated a space is, but that it doesn't have a direct effect on cognition by itself. The original study may have been picking up some other pollutant where Scott's didn't.
Personal anecdote - I was taking a 3 hour exam in a lecture hall with ~100 other people with no circulating air. I felt like falling asleep due to how stuffy the air was, but others seemed like they were doing fine.
My family used to play this all the time as the game Boggle. Fantastic game, I crush all my friends mercilessly if they are foolish enough to challenge me, who has been forged in the fires of family game night for years.
So, as a submariner, I think I should point out that CO2 levels are very tightly controlled with atmosphere control equipment on board. It’s something that the OOD reports to the Captain every watch, and there is monitoring equipment in every space. Nonetheless, there are certainly higher levels of CO2 and lower levels of O2 on board than there are outside the boat - and every sailor on board can tell that it effects their mental abilities. It’s hard to describe, but the feeling when you come up and equalize atmospheres with the outside world and how that effects your mind is very noticeable. Now we certainly do very hard work very well, like you said, but mental effects from lower CO2 are real. Nonetheless, that’s obviously both anecdotal and referring to much greater changes in CO2 level than you are talking about, so it may not be a useful data point.
This just seems a priori kind of unlikely. 445ppm CO2 is 0.34 mmHg, and even 3000ppm is only 2.3 mmHg. Meanwhile, as far as I know pCO2 in tissues is around 40 mmHg, so even the highest concentration of CO2 you tried is 1/20 the normal concentration in tissues.
Amazon not only sells CO2 monitors, but also CO2 generators. They seem to work on the principle of adding citric acid to sodium bicarbonate. Of course, any acid, like the acetic acid in vinegar, should work. There must be some ppm of CO2 which has an effect on cognition. This would be approaching the problem from the other end. With appropriate caution, of course.
Also, increased partial pressure of CO2 should drive an increased respiratory rate to clear the CO2 by the usual physiological mechanisms of negative feedback. Measurements should show that respiratory rate correlates well ppm CO2. And at some ppm of CO2 the compensatory respiratory rate should be overcome.
Additionally, CO2 accumulation in the blood lowers the pH, (acidosis), which probably mediates the effects on concentration, headaches, etc.,
Actually, you want to be a little careful with that. When CO2 levels rise high enough to be unquestionably dangerous, they are dangerous very fast, e.g.:
"CO2 levels of more than 30% act rapidly leading to loss of consciousness in seconds. This would explain why victims of accidental intoxications often do not act to resolve the situation (open a door, etc.)"
It would definitely take work to get CO2 levels that high, but it could be done in confined spaces if you were fairly careless. That article mentions that an OSHA study in 2015 estimated about 90 deaths per year in the US by CO2 asphyxiation in confined spaces, with a remarkable 2/3 of those deaths being rescuers.
Nice post! Does anything else about air quality affect cognition? I would imagine at the very least it's harder to get work done if it's 50 or 90 degrees Fahrenheit in a room, and probably hard if the air is forest fire level. I often do feel that I think better outside, but it may just be that I think better standing up or walking. Indeed, at least stereotypically pacing back and forth is supposed to be good for pondering things and I feel there is some truth in this; it matches my own experience (not sure of the mechanism). Likewise changing environments may lead to coming up with new ideas. I've had experiences where I was stuck on something and then as soon as I've walked away a new idea comes up (lots of mathematicians do too). Still I don't know if this is about cognition rather than some sort of change in something. Of course, my feelings also may not reflect reality, but I think there's some truth to the idea that changing the environment leads to new ideas.
This might be harder to test. Especially since coming up with new ideas is very different from finding words. But I would imagine most SSC posts don't come to you while you're sitting down at your desks and while you are walking around. And also you find it hard to focus when there is someone else in the room (you said this in a previous SSC post at some point).
Sorry if this became a little incoherent but the general point is that at least my feeling is that there are some environmental factors that do not really affect cognition, and I'm curious which one it is. My suspicion is that there is something to the fact that a change of environment can be good for cognition, which is why so many scientists like walking. I also would guess that this also lends plausibility to the CO2 thing; it's a proposed mechanism for this because they get out of the high CO2-office. It's worth noting whether this effect is real in the first place, and what causes it anyway, if not the CO2.
Well think about where our ancestors spent most of their nights. Tiny tent/igloo/hut with all family members present. Longhouse with an entire tribe. Most of the year, all of the above had a fire lit and minimal ventilation in order to preserve heat. Imagine how high the CO2 levels were, and how low the O2 levels. Yet the ancestors of whoever is alive today thrived in that environment.
High-CO2-impares-cognition hypothesis makes a lot of sense to me, so here are two ways to save it:
1) Your performance in WordTwist is bounded in the type of intelligence CO2 is affecting: say, you play as good as you possibly can at 105 IQ, and CO2 causes your IQ to fluctuate 130 to 110 IQ. Variance in score is explained by other factors - experience, motivation, attention, and so on. Seems very plausible - the game is not that conceptually hard.
2) CO2 affects intelligence in part through motivation (I know I don't want to do anything hard if I can barely breathe!). So then you were feeling very bad (from high CO2 and other factors), you wouldn't want to play at all, and when you were feeling smart, energized and motivated (from low CO2 and other factors), you would do something more productive. If motivation and ability is correlated, and you only want to play at certain motivation, your score stays constant even if your ability is changing.
Hey Scott, could you post scatter plots of all your data here, correlation numbers are really not very informative at all - you can have low correlation despite a very obvious effect, and high correlation when the data clearly looks like a random cloud.
I bought a MH-Z19B co2 sensor off ebay a while back to wire up to a pi/arduino and save some $ from buying a whole meter, but not too shockingly never got to wiring it up. This might be the inspiration I needed for it!
If I do that and figure out a testing scheme, I'll make sure to come back and post a prereg of some sort.
Scott, you posted this essay within three hours of someone mentioning the Healthygamer Youtube / Twitch channel (on video game addiction) in the current Open Thread. I think you might be kabbalistically required to give it a look.
5-10 times 3 minutes might not be long enough duration. Do you have data available?
In all levels of school and university, the high CO2 levels causing impaired capability to participate was considered absolutely obvious consequence of spending prolonged time in poorly ventilated classrooms and lecture halls, which are many and numerous in Finland. I don't know if we ever become stupider, but after 1.5 hours, bad quality of air is obvious and people would start feeling sleepy. This was commonly attributed to CO2 from bad ventilation.
Abstract of the article by Zhang et al 2016 you linked to suggests that the effects are not due to CO2 but bioeffluents.
And no, it is not about the boring lectures. I attended meetings of SF club that has poorly ventilated rooms in the cellar of student housing. All activities are fun. Too many people too long time, many people claim it is noticeable.
However, here is another stupid question. Is the impairment effect about amount of CO2 in air or amount of oxygen in the air?
Air contains both, but near 80% of it is nitrogen. In closed space, oxygen is breathed in and CO2 is breathed out, the amount of nitrogen remains the same. To my knowledge, it is inert and uninteresting thing for all relevant parts the human biology. I don't have access to Zhang et al 2016, but in their experiment where they did not find CO2 effect, they added CO2 artificially into air that otherwise circulated freely. I wonder if introduced CO2 displaced nitrogen and oxygen at equal rates, that is, much more nitrogen than oxygen, but amount of oxygen remained much higher compared to enclosed space where humans inside convert oxygen to CO2.
"I wonder if introduced CO2 displaced nitrogen and oxygen at equal rates, that is, much more nitrogen than oxygen, but amount of oxygen remained much higher compared to enclosed space where humans inside convert oxygen to CO2."
Thinking about this more, it wouldn't necessarily explain the lack of effect in Scott's experiment. Plausible more people in the room -> both higher CO2 level and less oxygen.
We're talking about CO2 levels of up to about 3000 ppm at high concentrations. That's 0.3% of the air. Atmospheric O2 levels are about 20.9 % at sea level, and seem to decline to about 20.1% at 1000 ft elevation. So I would be very surprised if displacement of oxygen by CO2 were a significant factor in this common experience of "bad air".
I can absolutely confirm KPier's observations about chess (I play blitz games at chess.com) - there are clearly good days and bad days, they are mostly unpredictable (ie. unrelated to sleep deprivation etc.) and the difference between them is unexpectedly large
Registering a prediction before I read your results: it'll be one of those maddening things where there's an effect but it's not significant and it's much smaller than the previous studies, so you decline to draw a conclusion other than "so, maybe?". I'm four paragraphs in and the admission that you're playing the game because you wanna rather than for the experiment, while reasonable as a life choice, smells like a potentially strong confounder and I had to pause to think about why.
I believe what I'm anticipating based on own gameplay habits is a self-selection effect where the more up to cognitive tasks you're feeling, the more you play the game. Then if CO2 is only one is several factors in your congnitive state, even if it's the largest one, I would expect no correlation at all - your high-CO2 gaming will occur when other factors combine to partly, if not totally, cancel it out.
One thing to check is if you're actually measuring your CO2 level at all. Most of these cheap meters just auto-calibrate such that whatever the lowest physical concentration they see over a 24 hour period is set to a given internal concentration, usually around 300 ppm.
That said, you'd probably still see large swings. Probably.
I just wanted to express my admiration for the fact that you wanted to replicate this study on your own, and went to such lengths to do it. This is a good win for the scientific method.
I think for certain intellectual activities, there's an optimal level of intelligence. If you're more intelligent then that, you'll want to do something else. Magnus Carlsen made this point about chess[1]. I assume it applies to the wordgame you play as well. This would mean that when you had a low CO2 concentration in your room, but still wanted to play the game, there must have been something else negatively affecting your intelligence at that moment, otherwise you wouldn't have wanted to play at all.
Could level scaling in the game confound any attempt to find (any) correlation (at all)?
I don't know this game, so this might not be applicable at all. Just throwing it out there.
But I'm guessing you would play many rounds in a single sitting, presumably at the same level of CO2. If you do well/not-well in one round, will the game compensate by giving you a harder/easier match next round? I'm sure there are game designers in the audience who can explain better than me why this is good/bad game design, but either way, many games do something to this.
One way to try to see this could be to use only the first game in a sitting as data. I don't have the skills or time to do this properly from here unfortunately.
Hypothesis for confounder: your brain switches signalling strategies between ones that are more or less CO2-tolerant in a predictive way. It uses your circadian rhythm, metabolic rate, etc. as independent variables, and observed CO2 as the dependent variable for this prediction. Your brain manages to function fine (at least in the short term) in both “modes.”
Then, it would only be when your brain mispredicts CO2 and stays in the CO2-intolerant mode while actually experiencing high CO2 load, that you’d see the conditions from these studies.
Under this hypothesis, your brain would probably re-predict easily enough when going from outside to inside, or when opening/shutting windows; so I’d guess you’d have to do something more like “releasing a bunch of CO2 into the room without otherwise changing AQ” to trigger the effect. (Which, if I understand correctly, is similar to what the studies do.)
This makes way too much of a very flimsy and poorly conceived informal experiment, if I understand it correctly. Start with an environment where CO2 levels fluctuate quite a bit (from about 500 ppm to about 3200 ppm), measure CO2 level just after playing WordTwist. Calculate whether there’s any correlation between CO2 level at end of game and score.
In other words, the only thing that was tested is, correlation between momentary exposure to CO2 to game score. No attempt to measure effects of continuous prolonged exposure is mentioned. Yet results are used to challenge more formal experiments that are measuring prolonged CO2 exposure, such as exposure of 8 hours.
This WordTwist experiment is of no significance, unless you hypothesize that the significant effects of exposure to high levels of CO2 is immediate and very transitory. I don’t think there’s any good reason to think that, and it’s certainly not what the studies that are being challenged presume.
It’s an important health and social issue as governments have been supporting or requiring, as a conservation measure, tighter construction to seal in heat, as well as indoor pollutants like CO2, which can easily build up, for example, overnight in a tightly sealed bedroom. The concern is, effects after such a prolonged exposure. If you want to do a realistic informal experiment, try, spending 8 hours in a room where CO2 builds up to say 2000 ppm or higher the whole time. (Well, don’t actually try this by hanging out in a poorly ventilated room full of people until after the pandemic, please.) That would be equivalent to the studies that are in question here. I’d even be willing to bet, your WordTwist scores would show a downward trend after several hours in this scenario.
The posting also asserts that cognitive impairment due to CO2 buildup in the ranges we’re discussion (like, 1500 - 3000 ppm) isn’t a concern on submarines, for example. There’s a comment below from a veteran Navy submariner who reports cognitive issues from CO2 buildup was indeed a big concern in practice. Though, as you note, there’s a Navy study claiming to have found no cognitive effect from even much higher levels of CO2 exposure — but the Navy only tested cognitive effects after exposure to high CO2 levels for 45 minutes, that is, intermittent exposure. Might make sense to test that if as the comment below suggests, steps are already in place to prevent longer term crew exposure to excessive CO2.
If you’re worried about whether a momentary variation in CO2 exposure is going to fry your brain, this WordTwist test should be reassuring, though I’d wonder why you’re worried about that. If you’re worried about a real problem — how conservation efforts may create indoor buildup of CO2 resulting in longer term exposure, this WordTwist test is a straw man.
I moved to a new house and bought 5 CO2 and particle sensors (nerd fact: mh-z19 and sds11 with esphome, 25$/each) to track different rooms. All are recording readings every 10 mins (nerd fact: into my homeassistant and then influx for longterm storage). I also got many cheap (10$/pc) humidity/pressure/temp sensors doing the same and I plan to monitor this for years for the whole 6 people household. Can't wait to see how it (doesn't) correlate with e.g. number of tasks etc.
I did something like this once. I wanted to try clonazepam to see if it would help with some work issues I was having, but I was worried about the cognitive side effects. I coded up the reverse-digit-span test from the Wechsler test as a quick Python program and took it regularly at various times for a week (it takes about two minutes), then started the clonazepam.
What I found was that clonazepam at the dose I was taking had no detectable effect, but there was a huge, massive, consistent improvement in my reverse digit span from taking a half-hour nap in the afternoon.
The thing I have the biggest trouble with is not short-term memory but executive function. I keep meaning to code up the trail-making A and B tests (the reverse digit span thing is a couple hundred lines of code) but I haven't gotten around to it yet.
isn't the way you happen to be breathing during each game a significant confounder here?
I recommend reading "Breath" by James Nestor for some insight into the variability and health effects of different breathing patterns. maybe using something like The Breathing App before each game in the experiment would help?
I am willing to do this and put in a minor amount of effort into a cognitive test (<2 minutes) once when I wake up and once when I fall asleep for about a month if someone is willing to pay for the monitor. Someone will also need to tell me which cognitive test to take.
By coincidence, I recently did something very similar (but did not record my results, so everything that follows is impression based.) I played several hundred games of a very similar word game on a 4x4 grid [Boggle with Friends], and found that I performed markedly worse under some pretty predictable circumstances:
- Before coffee
- Sleep deprived
- Drunk
- Drowsy antihistamines (cetirazine)
The interesting part is that board composition (availability of high-scoring words and the supply of T's, E's, S's, R's etc.) had a much higher impact on the score than any of those factors, and that makes me think that boggle-like word games might not be a good metric for measuring variations in cognition.
Also interesting was the fact that the following conditions didn't have a pronounced enough impact for me to remark on them at the time:
- Low to moderate alcohol intake
- Environmental distractions (not touch based)
- Non-drowsy antihistamines (fexofenadine)
- Hunger
- Time of day
And that the following DID have a pronounced enough impact to note at the time:
- Sitting position
- Touch-based environmental distractions
- Attractiveness of opponent (?!)
- Perception of time pressure*
*Notable because the game clock is always 90 seconds, no matter whether or not you feel like you're jamming the session in between two other tasks.
Don't know if this is a "really good experiment" but I'm interested in testing memory, logic, and creative thinking in the form of solving sunday nyt crossword puzzles (i recently bought a book of 500 of them) in a number of fixed environments (eg desk at home, sofa at home, outdoor patio, coffeeshop, park)
I am interested in tracking co2, particulate matter, temperature, humidity, light (brightness, natural vs artificial), visual stimuli (eg plants vs people vs nothing biological, clean vs cluttered, movement vs still), sleep quality/quantity, diet/weight, and time of day (in addition to solving "performance" obviously).
If the meters in the amazon list are purchased for me, I commit to solving at least 1 puzzle a day until the book is completed, share the results with Scott (and anyone who's interested), and to ship the equipment to whomever Scott directs me to upon completion of the study (I'm a minimalist and object more to there being more "junk" in my life/the world than the actual cost of the equipment).
Addendum: I would also track subjective physical/exercise activity, altitude and blood oxygenation (I own a pulse oximeter and its low utility makes me highly resistant towards buying the other gear).
I intend to travel quite a bit on extended road/camping/rv trips over the "study period" and think the data could be pretty interesting.
If 3200 ppm causes headaches and nausea, consider me surprised that it has no effect on cognition. Did you plot your data to see if there's an effect in the highest range?
Quite simply the reason you failed to notice any difference in performance related to CO2 level is because you are not a plant.
The experts say the Earth is warming due to increased CO2 levels caused by humans and the graphs show a bump in the recent past although, nevertheless, we are technically still in an ice age, which has been going on for 2.6 million years, with 100,000 year ups and downs.
Currently there's a human caused CO2 increase. Sixty years ago CO2 was only 380 ppm, whereas now it is 413.93 ppm. When CO2 increases, plants thrive and the Earth greens.
Plants inside a greenhouse prefer a ppm of 900, which boosts plant growth and health from 25% to 100% depending on the plant. As for humans the current 413 ppm is no problem. The CDC says you need 100,000 ppm of CO2 to die and other sources say 40,000 ppm is unhealthy for 8 hours. A submarine's scrubber turns on at 8,000 ppm.
Earth has had much higher CO2 levels than the current 413 ppm and our ancestors apparently thrived. 150 million years ago it was at 3,000 ppm and 6,000 ppm farther back. The cyclical CO2 increase is caused by the oceans' warming which releases a bit of the 93% they contain.
Current climate science may be putting the cart before the horse and one might question the causal direction that 97% of professional climate scientists agree on. But experts are purveyors of groupthink - it's how they make their money and continue their funding; they should not be trusted to give a definitive answer.
In the 1980's and in 2005 Phillip Tetlock studied experts' predictions and found they were no better connecting with reality than anyone else. One might spin this result by saying that we live in a random universe and the experts correctly identified and properly proved that fact.
Thus, regarding CO2, there's a 50% chance we really need to produce more of it. Carbon credits should be carbon debits and we must increase our individual carbon footprint to save the planet. But we could be wrong. To best determine the correct answer what's needed is a monkey throwing darts.
Would you expect office workers to also do that?
> body instantly abandons all normal priorities and concentrates only on detoxin this
Wow. How is that orchestrated in the body? How does the human body get rid of the toxin?
Is this a similar effect to people developing tolerance to low oxygen/low air pressure conditions? Or maybe actually exactly the same effect (i.e., increased red blood cells in people who are in low oxygen environments for long periods of time?)
How can you "metabolise CO2 better to avoid acid buildup"? The human body doesn't have any non-acidic way to transport it.
As a rule, if the blood bicarbonate buffer shifts towards H+ (excess CO2) your medulla just cranks up your breathing so you blow off more CO2.
I am less impressed with your study than I am with you finding intrapsychically.
I didn't - at the end of the game it tells you all the words you missed.
Is the CO2 in my room really high (possible, even with windows wide open; thanks, wildfires!), or does that word not actually exist in the grid? There is no neighboring "C" and "H" pair on this board.
There's a c on the far right column, and an h to the top left of it, I thought diagonals didn't count at first so I was confused
Extremely cool. One possible way to salvage the original result is if there is a cumulative threshold effect: maybe you need to be in a low-carbon environment for a long time to see effects and if you regularly breathe fresh air your mental function is not affected. This wouldn't explain the submarine data point though
Unfortunately, some of the pro-CO2 papers claim acute effects. Like Satish's experiment used 1 hour of exposure pre-testing (https://ehp.niehs.nih.gov/doi/10.1289/ehp.1104789), and is one of the most-cited ones and has p-values out the wazoo.
The whole concept is silly, IMO. If something we did regularly was messing with our cognition it would be obvious.
I'm not sure it would be that obvious. Recent analysis of court records have shown that judges levelled more severe sentences against guilty offenders the lower their blood sugar, ie. that hungrier they got/closer to lunch, they became hangry. After their lunch their sentencing severity popped back up nearly where they were at the start of the day.
I suppose we'd have to survey lawyers if this was common knowledge, but my understanding is that it wasn't.
The hungry judges story has been debunked. See "Overlooked factors in the analysis of parole decisions". https://doi.org/10.1073/pnas.1110910108
It's an interesting example of a claim that spread through headline repetition, when anything other than headline repetition would have instantly discredited it. The effect size found in the study was a *one hundred percent reduction* in the likelihood of being acquitted as you get closer to lunch.
Without any knowledge of lawyers, I'd say they were aware that the guy scheduled immediately before lunch isn't going to be acquitted, and isn't going to take very long - but that they would have provided the same explanation as all subsequent discussion of the paper has, that the court schedule is not determined at random, and that guy was placed before lunch because he was going to be convicted in a short, predictable amount of time.
Looks like the authors defended it: https://www.pnas.org/content/108/42/E834?utm_source=TrendMD&utm_medium=cpc&utm_campaign=Proc_Natl_Acad_Sci_U_S_A_TrendMD_0
I found something else more recent which found the effect size to be overestimated* though because of the varying length of cases, instead of time of scheduling, so it may still be debunked. That's as far as I'm reading for now, but there's a lot to it. Thanks for sharing that, I hadn't thought to question it before.
But it is obvious: ‘we all know’ that hunger affects our mood. Just like all teachers know that kids are more rowdy just before lunch.
(Remember that “we all know” does not mean it is true.)
Also, it's a long way from "affects our mood" to "affects our mood in this particular way, under this particular circumstance, all other things being equal, x% of the time."
"Go get some fresh air to clear your head" is common advice, so there's definitely something obvious enough to be noticed by regular people.
Definitely worth investigating.
Sure - but like the 'hungry judges' story, the purported effect is just too big.
I recently (June) started monitoring AQ in my condo after reading similar studies, watching a presentation by DHH from Basecamp (MRqh8oLY7Ik) and trying to find ways to help my wife's migraines. I'm using the Awair AQM8002A.
Our overnight CO2 in our bedroom (shared with 3 children) went up as high as 1300ppm and rarely went under 900ppm any time during the day. I tried opening windows and I could push it down to 700 but as soon as I closed the window (since it's July in Houston) it shot right back up. The humidity however took much longer to recover.
We left for a month and it went down to the 400s (set the temperature to 78 instead of 71).
I just used Condensate Pan Treatment Tablets (Uric acid) two days ago and our overnight reading have gone down to high 700s (from 1300ppm). I've had a lot of trouble figuring out how such a large change in CO2 could have been caused by micro-organisms in the condensate pan... My first thought is faulty meter, which I haven't tested.
No way to know the effect on migraines as it's a long term occasional problem. My personal waking fatigue is much improved though.
How did kids 2 and 3 come about with you all sharing a room?
Asking the real questions
Inquiring minds want to know!
I'm still trying to figure out how we managed not to have any more.
A friend’s mom had 9 kids and her favorite line was, “I really gotta find out what’s causing it.”
https://youtu.be/fUspLVStPbk
"We have five children because We Do Not Want Six". - a now-disgraced comedian
"Why do I have fourteen cats? Because I had sixteen cats and two of them died." - Paula Poundstone
apparently some of the cheapo meters actually use VOCs as a proxy for CO2 levels
yeah this is quite common. OP should double-check that it's a real co2 detector (I think the good ones usually use IR)
Yup, mine's a good one. Well calibrated compared to industrial CO2 meters.
Is urban dictionary a valid source of words for the game? Narc…
Could there be factors other than CO2 that correspond to ventilation? Air flow, humidity, level of VOC, sound levels, etc.
I sleep much better with a fan. A study might show that people sleep better with fans running. Headline, “Studies show high air flow helps sleep.” Later we find out it’s the white noise that’s helping.
My CO2 meter tells me humidity, but I didn't think to record it. Guess I've got to play another 800 games now. Good thing I'm addicted.
You should go decide which environmental factors to record for each game in advance so that you don't need to play 800 new games every time someone brings up a random factor that is potentially correlated to intelligence.
Some factors that might be worth recording, in addition to co2 and humidity, in no particular order:
- How much you slept
- VOCs
- Ambient sound level
- Air flow
- Stress (as in how much other stuff you have to do soon)
- Sugar intake
- Lighting
- Screen time
- Commute length
- Obesity?
- Chewing gum
- Watching stupid entertainment shows
- Fluoride
- Meetings
- Secondhand smoke
- Whether you took Ambien/Xanax/etc
Also, random thought: If you don't want to play the word game yourself a bunch of times in order to test hypotheses about human cognition, you might want to ask (some of) us, the vast ACX readership, to do it on your behalf.
Recording for each individual game, how many games immediately prior in a session and how you scored, might not be a bad idea. I get into a zone with my certain computer game and my performance usually improves for a few games in a row, then drops and improves again. If you were playing one game a session you’d always be cold, but if you play in blocks this tendency to temporarily improve might obscure whatever the CO2 was doing to your scores in general. Given a repeated task in the presence of some high but more or less constant level, maybe the brain adapts to the situation at the time.
Also, given that there were other people in the room sometimes, the presence/absence of conversations and/or singing.
This is a great list of metrics, but I'm not sure how I'd measure some of them (e.g. air flow).
If you are modifying the experimental conditions anyway, and there are no budgetary constraints, I'd suggest following additions:
(1) Air oxygen meter.
(2) Get an app that beeps at regular times (or random times could work too) and asks you to record if you feel like playing or not and record the air metrics. Get automated sensor that record the metrics anyway, so it is less work. Maybe you know someone who can hook the sensors into an Arduino board / RasPi?
The previous data from the same location can also be quite useful.
Humidity could be, but I don't think it's likely that it causes the Dutch East India Company.
> I don't think it's likely that it causes the Dutch East India Company.
I also agree with you that humidity is not the ultimate cause of the existence of the Dutch East India Company, but I also think that the phrase "Dutch East India Company" in the above comment is a malapropism.
I think they're making fun of you and BronxZooCobra for using the acronym "VOCs" without ever explaining what it means, since VOC was also the acronym for the Dutch East India Company.
Google suggests "Volatile Organic Compounds". I assume it's something along the lines of pollen count.
Not pollen, more like butane, propane, off gassing from paints, fumes from pumping gas, etc.
You might find this amusing, if you haven't heard of it before: There's an urban legend that sleeping with a fan can kill you. https://en.wikipedia.org/wiki/Fan_death
(I also sleep with a fan. It hardly ever kills me.)
It can lead to career death for artists these days.
This is a great joke.
Some spaniard scientist seemed pretty alarmed by the effects of infrasound on airplane crews´ blood vessels at one fascia congress and advised to shun computer fans. But that was at a conference years ago and today I guess there's not much behind it.
My anecdotal evidence is that I feel better and more alert when I ventilate the car on long drives. But in car CO2 levels be much higher than a house or apartment. I can’t think of a good way to test this.
The carbon monoxide might be playing a role in the car situation.
I had wanted to measure this exact same thing - ever since I moved into my current flat I feel that I can work so much better with the windows open. And the price did stop me - I thought these monitors were around $10 when I got the idea (why do they cost so much?). So... if logistics of me being in the UK is not an issue, I'd be happy to test?
Send me an email at scott@slatestarcodex.com with your planned experimental protocol (doesn't have to be too official) and your address and we can talk about it.
If you're actually willing to buy the monitor for someone(s), I suspect you'll find plenty of video game streamers who would be willing to do this as a gimmick on their stream for some weeks, so then you'd just have to decide which game(s) you think are more "cognitition" focuses.
If any streamers are reading this and willing to work with me here, let me know.
I'm a streamer. https://www.twitch.tv/nebupookins I'm open to playing a wide variety of games on stream (except I don't like to play horror games). I've also bought two CO2 monitors after reading this post (not the one you recommended, a $30 one and a $50 one) with the expectation that if they both report the same reading, they're probably both accurate, but if they report different things, then one (or both) of them are inaccurate.
I'm really curious if you have some suggestions for experimental protocol.
Are you playing any competitive online games? Overwatch, League of Legends, Starcraft? Those usually have pretty detailed statistics for player performance. I play LoL and third party apps like Porofessor show me very detailed stats per match, from "teamfight participation %" to "creeps farmed per minute". But to avoid multiple comparison adjustments, I'd just stick with measuring competitive score progression. If you gain more points when playing in the fresh room than in the stuffy room, this is good evidence for the effect.
I'm streaming too sometimes, a mix of Russian and English streams, gaming and photo editing. https://www.twitch.tv/loweren
I used to play League, but my performance in League already has insanely large variance so not sure it'd be a great metric for me. E.g. creeps per minute varied by a factor of 18.5.
A lot of recent top-level chess has been sponsored by an air quality monitoring company who used this gimmick. Eg you can see it at 1h30m here: https://youtu.be/EmVuLy08Y4k. The commentators sometimes mention it too, in order to help the sponsor.
If any streamers are reading this: I would find this gimmick pretty entertaining, and would direct some money/views toward a streamer who did this in an at-least-moderately-careful way.
How would you make it fun without unblinding the gamer about CO2 levels? Or maybe blinding is asking too much since people can usually sense low oxygen anyway.
People can't sense low O2 - that's why simple asphyxiants like nitrogen and the noble gases can kill. What we sense is high CO2 directly.
(Simple asphyxiants are dangerous because breathing them doesn't give O2 but does remove CO2 - as such, you don't sense "bad air", you just pass out from lack of oxygen and then die.)
Fair point, I’d learned that and then forgot it. But yeah replace low O2 with high CO2 in my post.
Confused though, nitrogen and noble gases aren't converted to CO2 by the body right? Is this part of why CO is so bad? CO would be converted to CO2. Though CO binds to hemoglobin so strongly this effect is probably negligible for CO.
What's the mechanism for this exactly? How do we sense the high CO2 that makes us want to breathe out? CO2 is pretty inert. Is it through CO2 acidifying water? If I put another gas that acidifies water in the air does your body start to process it as CO2, and will you have trouble even if there is enough O2 in the air?
Lots of questions, but would like more explanation here!
I'm speaking about the case where you're breathing pure N2 (or another gas/mixture that while not toxic does not contain O2).
If you're breathing pure N2, you will indeed maintain low levels of CO2 and not be poisoned by it. However, you need a constant supply of O2 in order to live, and pure N2 by definition doesn't have any. As such, after a couple of minutes of breathing pure N2 you will pass out from lack of oxygen (and then die unless rescued).
This form of asphyxiation (no CO2 accumulation, just lack of O2) can only occur if you are breathing gas with low CO2 and also low O2 - this is highly unusual, because "bad air" breathed out by other animals will reach dangerous levels of CO2 long before the amount of O2 drops to insufficiency (40,000 ppm of CO2 is 100x current baseline/150x pre-industrial and immediately dangerous to human health, but that's only 4%; animals use something like 1-1.5x as much O2 as they produce CO2, but even a 6% drop in O2 from the 20% baseline is still 14% - more than half of it is still there). The usual cases for sudden anoxia are sudden decompression of a plane, huffing helium balloons, SCUBA accidents or improper storage of cryogenically-liquified gases - to put it mildly, not the conditions natural selection designed us for.
I'm not 100% sure of the way by which we sense high CO2. I do know that strongly-acidic gases like HCl, NO2 or SO3 are very dangerous to breathe, although this has more to do with direct acid damage to the lungs than with homeostatic errors.
Cool! What's the mechanism by which CO2 is dangerous? If I breathe in something is just like air except with the CO2 in place of the nitrogen, so the oxygen content is still good, what happens exactly? Acidification, some of reaction A + B <-> CO2 + X where the equilibrium is off, something else? Why would I die? Isn't CO2 a relatively inert gas?
It is one of the cases where evolution cuts corners that no good engineer would: it tests something easy (CO₂ → acidity was my guess too) that is strongly correlated *in nature* with the thing it needs to test (O₂).
I learned recently (although I would need confirmation) another example: water. We feel thirsty when we need water, and we get minerals at the same time. We do not have a feeling for the lack of these minerals. Therefore, drinking distilled water will sate our feeling of thirst, but will not supply our body all it needs.
CO is bad because it binds to the heme group in hemeglobin better than O2, so it will displace O2. That is, your red blood cells will absorb CO before they absorb O2. It also induces a conformational change in hemoglobin that inhibits oxygen unloading, so even the O2 that is left isn't delivered efficiently to tissues. CO2 doesn't bind to hemoglobin at all, it's quite a different looking molecule.
Yes, the main sensors controlling breathing in the brainstem sense H+, which increases with CO2 levels because dissolved CO2 largely forms carbonic acid by interaction with water. There are additional sensors that directly detect O2 levels also.
Acidifying blood per se is not thought to trigger increased breathing, because the detectors are actually sensitive to H+ in the cerebrospinal fluid, so something has to cross the blood-brain barrier (which H+ cannot) and *then* cause a decrease in pH. CO2 does that.
Very cool. So, is there any other gas that makes a weak acid and achieves the same effect? I'm thinking about SO2 (which is nontoxic right) for instance. SO2 also gives you the acidification thing? You don't have the A+B <-> X + SO2 thing though.
Haemoglobin does bind CO2, via the N-terminal amino group (which can be carbaminated). It's a minority of the blood's transport capability, and it certainly doesn't involve the haem site, but it's not correct to say Hb doesn't bind CO2.
Don't need to worry about which are more "cognition" focused. Cover a wide array of genres and stream types and see if any have a marked effect. If the effect is indeed specific to a certain kind of decision making, this could tease out what it is.
Personally, I feel like there *might* be an effect for tasks that depend heavily on what people call "emotional intelligence" (scare quotes), but not for anything else (unless you want to talk about long-term carbon dioxide overexposure or something, which is a different topic).
Or for things that are tedious.
That could work too, but I feel like that might go away for tasks you're familiar with, so it might be impossible or hard to test.
Yeah I'm wondering about coming up with new ideas vs repetitive things. Again this is because a lot of people come up with new ideas while outside walking, and my suspicion is that the mechanism for this is probably not CO2 but it's a reason the CO2 hypothesis sounded plausible to me. I also feel I can focus better when the window is open, but it could just be a placebo effect or it could be that the sound from outside is some sort of white noise, or that having some change in the background stimulates new ideas (maybe it's extra hard to work and come up with new ideas while staring at a white wall) or something else I don't know.
I think there is a lot of interesting stuff in this space that is chronically under-explored, and which needs more empirical data so we can all figure out how to work more efficiently. This seems like a pretty underfunded direction of research...
It would also be interesting to see if there is a "natural" way to split various human activities, and whether differential effects on how well we can do them from various outside influences can be used to classify them. So is there some way in which tasks A,B are more alike to each other than C,D, in that A,B both become harder under circumstance X but not C,D? It would also be interesting to look at the corresponding brain activity while doing all of these but this would be hard unless you had some kind of mobile pocket MRI machine.
Again I think there are a lot of interesting questions here. We should try to understand them better because the upside could be potentially ginormous.
Nice self-replication! And I'm pretty convinced by your successive ruling-out of explanations as to why the effect is real but you failed to detect it (e.g., controlling for temperature).
Sorry if you mentioned this and I missed it, but is there any chance that time (like "index" of the trial/game) was a factor? Like, maybe you got better at the game over each successive play, but CO2 levels were also positively correlated with time? In which case you might mask a "real" effect of CO2 if performance is driven by Time (+) and CO2 (–). As you say, to salvage the effect you'd need for these effects to be of roughly the same size, which is a stretch. But might be worth checking.
Another possibility would be that you're already so good at the game that you're essentially at ceiling, so there's just not enough variation in your performance to explain, even though you observe plenty of variation in CO2 levels.
I don't love this explanation because to take it seriously, you'd need an account where the influence of CO2 is dependent on prior ability in the task. Basically an interaction between Expertise and CO2 on Performance, such that High Expertise is somehow "immune" and Low Expertise is more susceptible to CO2 levels. I guess this is in principle possible, especially if the task is easy enough for people with high verbal reasoning/fluency. But: 1) it doesn't seem that intuitive to me; 2) if it is true it's a pretty big caveat for the initial results, and implies they must've sampled from Low Expertise population to have detected an effect (assuming they didn't regress out Expertise); and 3) now we're just adding epicycles.
If you mean getting better over successive games over the course of months, this is definitely happening, but CO2 isn't changing consistently (other than I guess in the global warming way, which is much slower). If you mean getting better over successive games in the course of a sitting, I have the data that would let me test for that but I haven't done it - just eyeballing it it doesn't look that way, nor does CO2 go up very much over the course of my sittings (I usually play only 3 or 4 games max in a sitting, which takes about 15 minutes, which isn't enough time for room CO2 to increase much)
I'm definitely not at the ceiling - my "score" (which I calculate artificially as score divided by average score for the board) has varied between 54% (my worst game) and 143% (my best) over the past month. The world record for most boards is often around 2x my own score.
Yeah I meant the former ("getting better over the course of months")––I just thought maybe if time was correlated with wildfire season or something, you might detect higher CO2 levels. But that makes sense! Seems like a solid non-replication.
“ Several parallel scenarios are available, allowing retesting individuals without bias due to experience and learning effects.”
This is a quote from the paper that Scott cited. I wondered the same thing as you; whether a lot of prior knowledge of the test affects performance enough to overcome any problems with CO2 levels
The lesswrong post linked about chess is nearly a good measure but is made much noisier by the confounding factor of "how good is the random anonymous person I happen to be playing today?". Happily chess furnishes a much better measure of day-to-day fluctuations in cognition, which would be to solve chess problems: positions (either taken from real games or artificially composed) in which there is a single, objectively correct answer to the question "What is the winning move here"? This also seems more like "decision-making" than finding as many words as you can in a word-search puzzle in a set period of time.
Does chess have a recognized solution to this? I don't know exactly how Elo ratings work, but it seems to involve adjusting for your opponents' skill (eg you get lots of points for beating someone better than you, and you lose very few points for losing to them). Would "average Elo points gained/lost per match" be a fair measure of how well you're doing that day? Do chess websites make that easy to track?
Oh sure, they can calculate and track this automatically, but a whole game of chess consists of so many decisions that you could feasibly lose to one person on one day because you made 40 great moves and then one dumb move (median cognitive strength: good!), and then beat someone else of exactly the same elo strength the next day by making 40 mediocre moves and then one more move that didn't lose (median cognitive strength: average!). The advantage of solving a single problem (ie choosing a single move by calculating a huge branching tree of variations) is that there is no time limit (it can take half an hour or more if you like) and it's a single decision, so it is arguably a better measure of your peak concentrated cognitive power at that time. (There remains inevitably some subjectivity in how to grade the difficulty of problems though.)
In fact ELO assumes that anyone can beat anyone, just with different probabilities.
Well sure, this is how it's possible to increase one's elo by beating someone with a higher rating.
Elo is good for this. Given two players' ratings, you can calculate the probability that one will beat the other. So rather than tracking win/loss or +/- points, you can track if you did better than expected or worse than expected and by how much.
Chess.com also provides an accuracy rating (how closely your moves matched the engine's recommendation). This could work too. The only problem is that if your opponent is much worse than you, it's generally really easy to get high accuracy because the best moves are obvious. But with a decent match-making system I wouldn't expect that to be much of a problem.
I disagree with the other person's puzzle suggestion; some puzzles are just harder than others, and unlike with opponents, there isn't really an objective way to account for that.
I forgot to add -- I've played around with chesscom's API and using it to track the rating info would be easy, but unfortunately you have to pay for accuracy info. You could conceivably track stuff like "number of blunders, mistakes, and inaccuracies per move" by hand for free though.
You'd use performance rating and this would be easy to track via information from a chess website.
If anyone does decide to use chess as a marker for cognition, I would recommend ignoring win/loss entirely and focusing entirely on moves. Use stockfish or something to compare scores before and after each move. Proportion of good moves or some other move based transform will be more robust to blunders (by you or your opponent) than game win/loss.
Any chance you can test the correlation against different CO2 sensor noise/bias models? I believe you that low readings happen when you'd expect and high readings happen when you'd expect, but I can imagine observing that even if (CO2 reading) = m*(True CO2 level) + b or even something higher-order.
I don't know what this means. If you think you can do it with the raw data, I can send you the raw data.
Nah, Kenny convinced me. Thanks for replying though :)
If X is correlated with a linear transformation of Y, then X is correlated with Y to precisely the same degree. If it's not a linear transformation, then the observed correlation could end up being larger or smaller, but to completely erase a correlation that is in fact strong would likely involve a weird transformation.
I did something similar playing Go against the computer when I worked long overnight shifts in an intellectually focus demanding job. Every few hours I'd play a few quick 9x9 games against the computer and decide based on my resulting rating how well I was mentally handling whatever level of fatigue I was in.
Scott: instead of doing univariate p-tests, why not fit a linear model where you add an extra term that depends on days since you started playing to model your skill increasing over time? I’m happy to fit the model for you if you feel comfortable sharing/uploading the data somewhere.
Done! See http://slatestarcodex.com/Stuff/CO2data.xlsx . You'll want to use "ME AS %" as the dependent variable, that's my score as a percent of the average score for each board. I didn't start recording some thing until later so those columns will be blank until I start. Let me know if you have any other questions.
Excellent! Will take a stab at it this evening and post the results asap.
Hi Scott: short answer - there's absolutely no statistical effect that I can see from any of the variables you specified.
Long answer: here is a COLAB that anyone can run with all your data so they can check my steps, I threw this together pretty quick: https://colab.research.google.com/drive/1R8KBD_3Wvh_0KmI3YMcdzkMve6JXeXJ6?usp=sharing - if you are interested in the summary, it's the final cell at the bottom. As you can see, the total R^2 is pathetic (0.13) and none of the covariates are statistically significant.
Oh, and the final model is: 'normalized_score ~ total_days + minute_of_day_sine + C(day_of_week) + CO2 + temp' - with the C(day_of_week) syntax specifying that I used a dummy encoding to capture any effects from day of week.
Really interesting to see this. I started a company to commercialize a CO2 capture material back in 2014. At the time, a lot of people I talked with were excited about this study, but I was always really suspicious of the effect size.
I was planning to kick of some self-experiments on cognition in a few weeks using chess puzzles, working memory, and math tests. I ordered a CO2 monitor and will add that in to the mix.
Anyone have any other interventions or supplements they think would be interesting to test?
Thanks for the preregistration!
Np. I will post my protocol and planned analysis before I start, similar to what I did for my sleep study (https://www.quantifieddiabetes.com/2021/07/melatonin-to-stay-asleep-longer.html).
The sleep study just finished and I’ll be unblinding the data and analyzing this weekend. That’s what got me thinking about what my next study would be.
I think that CO2 has a big impact on my productivity, because I seem to get distracted much easier and am less willing to do real work when I spend a longer time in a less ventilated room. (In fact, I am likely to play a game like WordTwist in that scenario :D) But I never measured this with a monitor, so I would be interested how well my notion of distracted-ness really correlates with CO2 levels. I just can't think of a good, unbiased, way of measuring it...
Whelp, found out that I suck at word games today
No, that game is just very responsive to practice. I started out getting about a third of average, and after playing it a few thousand times (I only measured carbon dioxide for 800, but played it much more), I'm slightly above average most of the time. Most of the people whose plays are going into that average have been playing it hundreds or thousands of times, and learned all the stupid meaningless words that it accepts as real (some of the most profitable words are short extremely obscure ones using common letters, like telt, lari, esne, and tael). You also need to learn the generation algorithm (about half the time, it inserts s a very long word usually ending in -ness or -tion, so you often want to search for those endings and then try to work back from there). With a little bit of addiction and wasting your potential as a human being, you can become as good as anyone else!
o7
I put significant probability on the hypothesis that cognition responds to fast changes in CO2 concentration, but not to absolute CO2 concentration; ie, if you go from outside to a stuffy room, you'll be mildly impaired for some duration, and if you go from a stuffy room to outside you'll maybe get a boost for some duration (or perhaps an opposite-direction effect which is also some sort of impairment), but if you wait awhile in either environment you'll return to baseline. This is what you'd expect given a homeostatic mechanism that isn't instant. This would reconcile the original results (where people go straight into a high-CO2 room, and do worse) and your results (where you let the CO2 concentration rise at a slow, natural rate).
That's a good point, and maybe I'll try testing it sometime.
This makes sense. Blood co2 levels are not primarily driven by air co2, and the health effects of low ventilation that building codes are concerned about use co2 as a proxy for accumulation of volatile organic compounds and other gases, not co2 itself. At least climate change isn’t making us co2 stupid!
I will consider testing it, mostly because the game is fun. But also does anyone have any other circumstances to test this with that could affect intelligence? “Before vs after strenuous exercise”, “while fasting vs while not”, “before vs after big meal”
Oooh what about high vs low air pollution days? Although id expect that to fail too, the negative effects of partial combustion. product pollution is probably longer term. Many possibilities
Wouldn't CO2 also be a proxy for VOCs in my experiments, ruling out that VOCs cause these kinds of cognitive problems either?
Yeah, and the effects of VOCs aren’t short term cognitive ones. It is in the context of a rebuttal to the claim that co2 bad mask bad being backed by reference to the permissible co2 concentrations in buildings
*aren’t generally short term cognitive ones
Within one building, CO2 should be a good proxy for VOCs, but it's possible that either the low CO2 building or the high CO2 building happened to have a really gassy carpet or paint or something.
Not necessarily. Within a single room, local sources and sinks will predominate. For design purposes, or over buildings with tens of thousands of square feet, CO2 <-> VOC is a good (or at least reasonable) proxy.
Did you happen to experience the other negative symptoms associated with high carbon dioxide at the times in which your monitor recorded very high levels?
No.
This post needs more scatter plot!
For the love of Christ, Scott, the game is called Boggle. "WordTwist" isn't the name of a game, it's just an end-run around a potential lawsuit.
I agree they're basically the same, but they use different dictionaries, different letter frequencies, and different generation algorithms (Boggle's is random, Wordtwist's isn't) to the point where skill at one doesn't carry over to skill at the other very well, and in some cases you need different strategies.
OK dammit, that's a pretty convincing response :)
I thought I was decent at scrabble and boggle but I went and played WordTwist and felt like pond scum with 50 versus an average of 200
As Scott says above, you just need to practice. I started at 1/4 to 1/3 average on fives too (although I beat average on 4x4 with clever tactics)
doubled my scores on day 2, roughly. Mostly by speculatively trying 4 letter combos that sound like they might be a word, and focusing on vowels and suffixes.
Same! I’m now at 1/2 ish average. Spamming potential rare archaic words and trying to identify backbones that work for many words.
I don't think I have a very low credence on substantial individual differences on this. If nothing else, I get the impression that there's a lot of individual variation in how unpleasant it is to be in a high-CO2 environment. Still, for some people to have a large effect and some people to have zero effect would be somewhat surprising.
As a former submariner, I don't recall noticing any effects from varying CO2 levels. I do remember noticing huge effects from varying Oxygen levels as the concentration was adjusted. Lower levels and everyone was tired all the time. Higher levels it was harder to sleep and when I did, I would have bizarre dreams. Maybe there was an effect from CO2 but it wasn't noticeable to the larger effect of oxygen.
Was the harder to sleep accompanied by having more energy in a good way?
In a good way, in that you didn't feel tired all the time, but also this was sometimes a bad thing when all you wanted to do was sleep between your shifts on watch and couldn't "time travel" as it was called. There wasn't a lot to do when you weren't actively on watch, so it could be hard to pass the time.
Was the O2/CO2 varied on purpose, or was it a side effect of other things?
I think the CO2 was a side effect of crew activity, but O2 was controlled. I don't know this for certain as I was just a worker in the engine room side of things, but it was said they would keep O2 levels lower for periods when we were on mission so that more people would stay in bed and make less noise. The level differences were only noticeable on extended journeys when the recirculation / ventilation at periscope depth was not used as to remain hidden.
Is it consistent with your data that while CO2 has no effect on your performance *given that you decided to play the game*, it may affect whether you're up to the task of doing something vaguely cognitively taxing? In submarines and spaceships, you're in an environment that forces you to do things or else you die/are Severely Talked To by your superiors in an extremely strict social hierarchy; in an experimental room, it doesn't really matter whether you focus or not.
The game is called Boggle, tho that usually refers to the 4x4 version. I was addicted to it in grad school, but frustrated that the boards sometimes had only 3- and 4-letter words, hardly worth the effort of finding. So I wrote a program to generate Boggle boards with more and longer words in them.
First I tried making the probability of a letter showing up proportional to its frequency in English, but that resulted in boards where you could find TENT 3 different ways. Then I tried using a genetic algorithm that scored boards according to the frequency of the trigrams found in them, but that was even worse--it could fill the board with regular tilings of T ,H, and E.
Eventually I had the program count the dictionary words found in each board, and use that as the GA score. That worked well enough to run smack into premature convergence, which turned out to be so disturbingly powerful that it radically changed my views on the impact of technology on society.
(Now I think that rapid communications with high connectivity has taken us into a regime where we're no longer capable of evolving quality cultural artifacts / designs. Modern art was perhaps the first case of cultural premature convergence due to tech--an artistic ideology that spread like fire (their analogy, not mine) along the new rail networks linking Europe, and completely displaced all others in institutional art when it was still half-raw. I think premature convergence also contributed to the decline in innovation in pop music since 1990.)
For your purposes, you'd like to reduce the variance of the max possible score on each board. You could try giving yourself an adjusted score of (your score) / (max score). But I don't trust the scoring system to scale the right way with word length so that this adjusted score would have the same expected value on words of all lengths. Words like "intrapsychically" could mess with your adjusted score. It might be better to write a wrapper around the board-generating program that rejected boards with an unusual distribution of word lengths.
I use max score / average score of other people who have played the board. Usually there are about 20 previous players so there's not too much variance caused by the last player being really bad/good.
Can you explain more of what you mean by "That worked well enough to run smack into premature convergence, which turned out to be so disturbingly powerful that it radically changed my views on the impact of technology on society."?
John Holland's 1975 book /Adaptation in Natural & Artificial Systems/ proved the Schema Theorem of genetic algorithms, which guarantees that, over time, the "schemas" (genetic building blocks) with the best scores will be present in an exponentially increasing number population members. This means that the top-scoring genome is very likely to get fixated in the population very quickly once it has a foothold of just a few organisms.
This would be great if the fitness landscape were pretty smooth and lacked poor local maxima. But if it did, you probably wouldn't be using a genetic algorithm, which is useful largely because it helps organisms escape local maxima (without using randomness just to add noise, like most techniques do--it instead uses its random numbers to recombine partial solutions which each have high scores).
This would (I think) be okay if your genome were so high-dimensional (e.g., human genome, 3x10^9 base pairs) that each individual locus can have only a tiny effect on the score, and even chromosomal crossovers are likely to have small impacts on the score because they contain so many different loci that their scores will have a normal distribution with a very small variance.
But my 4x4 Boggle genomes were just 16-dimensional. So a single mutation could increase one genome's score significantly, and that genome would take over the entire population in a few generations. The population has then converged prematurely (before you ran all the generations you wanted to) on a single (crappy) solution.
I tried different ways of staving off premature convergence:
- increasing the mutation rate
- increasing the population size
- using log(score) instead of raw score
- increasing the mutation rate as the score increase decreased (this is the opposite of simulated annealing!)
Nothing worked until I tried dividing up the population into different groups, and severely limiting interbreeding between groups. For evolution to continue long enough to find good solutions, the population needed safe local niches where new genomes could arise without being instantly stamped out by the one dominant genome.
As I went around thinking about premature convergence, I started noticing it all around me. Starbucks. Tolkien imitators. The rise of a single national Native American accent. Plotless literary short stories about alienated characters who fail to resolve their existential crisis. All 1990s pop music.
Worse still, tech allowed perfect reproductions of books, music, and videos. In the 19th century, music and drama had to be live, and so nobody could capture more than a tiny fraction of the global market. 20 years after the introduction of the gramophone, an article in the March 25, 1908 /Musical Courier/ said that 3,000 professional pianists were starving in Vienna. Today there are perhaps only 5,000 professional pianists, starving or otherwise, in the entire US (https://www.zippia.com/pianist-jobs/demographics). In the 20th century, popularity shifted to a power-law distribution, so the top 20 artists in popular art forms make more money than the bottom 100,000 (I have looked at actual numbers in the past, but may be misremembering by an order of magnitude). Music and acting aren't professions anymore; they're lotteries.
Change and variety have dropped precipitously. English-language pop music changed more from 1964 (Elvis & the Everly Brothers) to 1968 (Pink Floyd & Jimi Hendrix) than it has in the 52 years since (and I'm not even going to qualify that with "might have" or "by some measures"). Literary novels became indistinguishable and boring. So did SF & F. I read My Little Pony fan-fiction now, not because I'm especially fond of My Little Pony, but because I think it's better fiction, and far more diverse in style and perspective.
I should elaborate on the sudden change in pop music in the 1960s, which came just /after/ a new technology (FM radio) started communicating hi-fidelity stereo music clear across the country. AM radio was too noisy to play Jimi Hendrix or Led Zeppelin. The first FM stereo stations were in 1961 (https://en.wikipedia.org/wiki/Radio_in_the_United_States); the first FM car radios were sold in 1963 (https://exhibits.library.gsu.edu/current/exhibits/show/georgiaradio/radio1960s/fmradio).
I think that the pattern seen when new technology suddenly connects previously-isolated communities isn't a gradual slowing-down of cultural evolution. First there's a sudden burst of innovation when everyone is first exposed to everyone else's ideas. That innovation slows down as everyone's experience becomes more and more similar.
A completely different application of premature convergence is in education, where we introduce it deliberately. Most Americans have always, for some reason, thought that ideally every kid in America should have the same education. Evolutionary theory suggests this is a terrible idea, culturally and economically.
Really interesting ideas! Though I'll point out the rise to dominance of hip-hop/rap as a significant change in popular music since the 80s
Yeah, culture is more complex than Boggle. I think premature convergence is a pattern or tendency, not a Law of History. Also my example of FM radio is sketchy, because shouldn't something similar have happened after the introduction of records in 1889, and after the formation of radio networks in the 1920s? (Maybe it did? I have no idea.)
It seems to me your characterizations of the cultural convergences as *premature* relies on your value judgments. Maybe the fitness landscape in these cases is quite smooth, and the evolution converges rapidly to what is indeed the global optimum.
That's a reasonable hypothesis, but the change in the distribution of popularity in culture (from maintaining population diversity, to being taken over by a power-law distribution of just a few big winners like Starbucks or Beyonce') matches the change in the distribution of genomes (from maintaining population diversity, to being taken over by a few, and then just one, genome) in premature convergence to a lousy global optimum. Culture with high-tech communication and reproduction seems less-able to sustain the population diversity needed for continued evolution.
This is clear in the case of language. The change in English from Beowulf to Chaucer (< 400 years) was huge compared to the change from Chaucer to Elizabethan English (< 200 years), which was huge compared to the change from Elizabethan English to the present (> 400 years). But then language is something that governments WANT to be standardized rather than optimized.
(Also, the Beowulf-to-Chaucer change was mostly driven by a sudden breakdown of linguistic isolation in 1066.)
Statistics on the changes in the distribution of years-to-bankruptcy, over the past 150 years, might be informative: are the big near-monopolies living longer? (My impression is no.) Or maybe a measure of product diversity would be better. But either approach would be plagued by complications due to the (human) population increase over that time, the shift to globalization, the shift to electronic media and social networks, and changes in regulation.
I would like to test the idea, but don't have any good idea right now how to do it.
Hmm. I'm not seeing how you the dynamics of the approach to equilibrium would tell you anything about whether it's a local equilibrium in a rough landscape or the global equilibrium in a smooth. I mean, if there *is* such a clue the protein-folding people urgently want to hear about it ha ha.
I think the only thing that would be solid evidence that it was a local equilibrium thing was if, after some time, the system (pop culture, language, whatever) lurched out of its apparent equilibrium and found some better one -- then we could conclude, post-factor, that the first was a local equilibrium and the fitness landscape was rough.
The inherent problem with this, though, is that we think the fitness landscape is *also* constantly changing, id est there isn't any absolute objective standard for the quality of pop music, it's a question of what the musical tastes of the moment might be. I think we can also make a good argument that the fitness landscape for language also changes all the time (e.g. through population migration, new technology, communication changes).
Using the phrase "solid evidence" here is problematic. There's a lot of territory between "not telling you anything" and "solid evidence". The dynamics of the approach to equilibrium tell me a lot in the case of Boggle board evolution, because I studied hundreds of cases, and the score of the final result was very, very strongly correlated with the number of generations before convergence. Is that proof, as in geometry? Not at all. But it's a whole lot of Bayesian bits of evidence.
The application to culture is speculative, and there are lots of other factors in play which would be difficult to factor out of an analysis. I'd be surprised if it were easy to identify many clear cases of premature convergence in culture. But I'd also be surprised if that dynamic turned out not to be a significant factor in cultural evolution.
An important line of supporting evidence which I haven't yet mentioned is the theory of island biogeography, which gives an equation relating the size of an island to the number of species its ecosystem can support at equilibrium, and a lot of empirical evidence that fits it well. The take-away is that evolution is much more powerful--you evolve many more species per square mile--if the territory is divided into lots of little areas with boundaries between them that greatly constrict gene flow. This seems to be due to the same dynamic I describe as "premature convergence". So it does appear to be a major--probably THE major--factor in evolution, which is a complex, real-world application. This makes it more likely that it's also a major factor in cultural evolution.
Blair Fix's essay 'The Half Life of a Spotify Hit' has some nice data and visualizations in this regard: https://economicsfromthetopdown.com/2021/06/18/the-half-life-of-a-spotify-hit/
You might also want to adjust the score by the frequency of each word in a text corpus. But again, it would be difficult to know /how/ to adjust the score, and the safer solution would be to reject boards with very rare high-point-value words.
My purely anecdotal self-observation is that my cognition powers drop dramatically when I'm tired, whether or not I realize that I'm tired. In fact, noticing that my cognition is impaired is a clue that I must be tired, even though I don't feel tired, so I lie down and am usually out like a light. This must far outweigh any effects of CO2 - unless the CO2 is making me sleepy, which it could be.
I am surprised by this (I was on team "CO2 adversely affects cognition" for a few years).
Interestingly there's a 2019 study (after my last round of CO2/cognition reading) that shows near zero effect of CO2 on cognition.
If anything, your results are more concordant with this study than any other study I know of about CO2/cognition.
https://www.nature.com/articles/s41526-019-0071-6
For me, the studies in this area are very frustrating to read, because they all track CO2 exposure for different time durations. For example, the Navy did a study (Rodeheffer) of exposure for 45 minutes (presumably because they had some reason for looking at brief rather than longer duration exposure), while the longest time of exposure studied seems to be 8 hours (Allen) (because it's the length of a standard workday, at least supposedly.) In one study (Satish) subjects were exposed 3 times in the same day and measurements done for each exposure (I don't understand why this protocol was chosen). It would be interesting to look at whether there was a pattern to results of first, second and third exposures but if this was in the paper I didn't see it.
Generally it looks like the longer the exposure period, the more likely the study is to see some effect, which might explain some of the differences results between studies that all seem to have been pretty carefully done. With this WordTwist informal experiment, however, we have no information for duration of exposure prior to the test, so I think the results can only be interpreted to negate cognitive effects of brief or momentary exposure to various levels of CO2 and can't be compared to exposure for longer periods of time. I don't know why it's meaningful to test momentary exposure to CO2 at these levels, but then again, I'm not understanding why any of the exposure times were chosen, except choosing 8 hours to replicate a standard workday.
Show your data. Show *all* your data. This would all fit so, so nicely into a single scatter plot, and I'd be about 5x more convinced if I could see it (especially if you can work in time data in a way that lets us know how your learning the game looks compared to other effects).
It's on the bottom as an xlsx file, you can do whatever you want with it.
I see a link in one of the comments, but I still can't find it in the original article? I feel like I'm missing something dumb here.
In any case, here are the plots I was thinking of. Nothing too fancy, I just like to see my data before I start looking at statistical analysis on it: https://colab.research.google.com/github/sclamons/ACX_CO2_poisoning/blob/main/CO2_poisoning.ipynb
I’m not a statistician, but isn’t p=0.97 a really high p value? My understanding is that if you have a completely uncorrelated thing your testing, your p value should be a uniformly distributed RV. So there’s like a 5% chance of p<0.05 and same for p>0.95. I guess this isn’t a real critique beyond “if I’m understanding the stats correctly this seems so high it makes me wonder if something’s up”
Ah, so it’s like how you need a high powered study to get a significant result when you’re dealing with a small effect size. Since this study has so many samples, if there actually is no effect it makes sense for the lack of an effect to show up more clearly?
Thanks, I forgot the difference between N=800 and N=30 lol
No, bored-anon is confused and you were correct initially.
P-values are derived by comparing the observed deviation to the calculated distribution of sample means. The predicted distribution of sample means becomes narrower with a larger sample, which causes a given deviation to result in a lower P-value.
The chance of P > 0.97 is indeed 3% regardless of the size of the sample - getting P = 0.97 with N = 800 involves a much smaller deviation than the deviation that would give P = 0.97 with N = 30.
Oh, okay. That’s why I said I don’t understand statistics at the end, lol. Deleted
It happens. No hard feelings.
But I could be totally wrong about the stats, it’s been 6 or seven years since I last did p-tests
I'm an environmental engineer, but I don't specialize in indoor air quality. It is not a climatic impossibility to have indoor CO2 levels < outside air CO2 levels. A local sink (most likely a houseplant) will easily accomplish that. Whether it's reasonable for your word gaming ecosystem is for you to determine.
I don't have a houseplant, but my window opens on a backyard containing many plants; would that do it?
Possible. You could test by placing your CO2 detector outside near the plants on a sunny day (and not during a drought - plants should not be water-stressed) and seeing what it reads. Ideally at some different times of day, too. If reading >> what you were seeing inside, it's not those plants causing the low reading. If reading is low vs typical atmospheric, doesn't prove it, but at least indicative.
It might autocalibrate to <400ppm though lol
Back of the envelope, sunlight is 1000W/m^2 and plants have a photosynthetic efficiency of at most 1%, so you're getting at most 10W/m^2 worth of CO2 converted to glucose. 10 joules = (10/4184) kcal = (10/4184/4) grams of glucose = (10/4184/4/180) moles of glucose, which requires (10/4184/4/180)*6 = 0.0000199 moles of CO2. So that many moles of CO2 are being absorbed per second by one square meter of the most efficient plants in direct sunlight. Meanwhile, one cubic meter of air is 1000 liters, and 22.4 liters at STP is one mole of air, or 44.64 moles per cubic meter. So if the plants were sealed in a 1 cubic meter greenhouse the CO2 PPM should drop by 10^6 * 0.0000199 / 44.64 = 0.446 per second. So it seems plausible that if you're in a very green rural area and the wind is blowing the right direction you could get a legitimate CO2 reading significantly below the worldwide average.
But google tells me "2000kcal / 24 hours in watts" is 96, so you need 10 square meters of the most efficient plants in direct sunlight just to keep up with the CO2 produced by one human. It seems very unlikely that this condition holds in any normal indoor area.
Could the plants reduce CO2 below outdoor levels while the house is empty? Suppose a typical house is 200m^2 with 2.5 meter ceilings = 500m^3 and has the equivalent of 1 m^2 of the most efficient plants in direct sunlight. Then the plants would reduce the CO2 PPM of the unoccupied house by 1/500 of the rate in that hypothetical greenhouse above, or about 3PPM per hour. It probably wouldn't get very far below the outdoor numbers before ventilation puts it in equilibrium.
Roughly half of all outdoor readings will be below worldwide average. There are diurnal and seasonal cycles due (at least in part) to plants, as well as human activity. And the human activity causes substantial local variation - this is why Mauna Kea measurements are preferred for tracking long-term global CO2 change. Here's a Chinese paper suggesting daily cycles ~30 ppm and seasonal cycles ~60 ppm in Beijing (note minimum is in summer): https://pubmed.ncbi.nlm.nih.gov/14551950/
Thus, the baseline to throw out data should at least be locally measured CO2 minima, not any global number.
I remember reading that houseplant VOC destruction, previously estimated to be very large, was in fact mismeasured, and more recent research had reduced the magnitude of the effect substantially. Can't find a citation with quick searching, though.
Assuming your math is correct, a houseplant may not be the only way to locally sink CO2. I'd have to think for a bit to say that there's nothing else. Globally, the other big sink is the ocean, but probably not enough water in a room to make a difference.
Yeah, the basic test for whether indoor plants matter is "are your plants adding mass in comparable quantity to the amount of food you eat?" (technically, the amount of [food - faeces] you metabolically process). Of course, plants only consume CO2 during the day and actually produce it at night, but if the mass gain of the plants is orders of magnitude less than how much food you eat then they can safely be ignored.
This is why motile plants don't exist; photosynthesis doesn't provide enough energy for locomotion. Animals get the required energy because they can tap a much larger area of photosynthetic activity than their own body area - a cow sustainably eating 6 hectares of pasture is (inefficiently) harvesting the entire photosynthetic production of that pasture.
How does that CO2 monitor compare with the "smart" one from Awair? Await keeps a record of the CO2 levels over time, so it might be easier to use in a study. I also don't know how accurate it is, but can tell that opening the window and the like affect the CO2 quality.
Andrew Gelman has a great term: “trying to weigh a feather in the pouch of a jumping kangaroo”. Sure there can be a ton of confounding variables, but if this were a meaningful effect (for certain definitions of “meaningful”) it wouldn’t require a research study that takes those into account to assess it.
Secondary problem is even if you did find an effect it doesn't account for adaptation effects. My prediction would be your body optimizes for a given CO2 level over some period. Consider climbers adjusting to higher elevation--takes a day or two.
It might apply more to motivation than calculation.
It's totally plausible that CO2 homeostasis mechanisms include throttling back the production of it, rather than just breathing more to get rid of it and excreting it in urine.
Throttling back production might involve the brain trying to avoid work when blood CO2 levels are elevated. But I don't know if 1000ppm is enough to have a measurable effect on blood CO2 levels.
https://pubmed.ncbi.nlm.nih.gov/31869112/
"Generally, under normal physiologic conditions, the value of PCO2 ranges between 35 to 45 mmHg, or 4.7 to 6.0 kPa"
At atmospheric pressure slightly above sea level, 100kPa, this is equivalent to 47,000-60,000 PPM. So an extra 1000 PPM in the air you're breathing is not likely to matter. It'll slow CO2 clearance by 2%
It's likely other factors that correlate with having windows open are causing the inconsistent effects in some of the studies effects: both lighting and odors can have emotional effects which might affect performance.
Also as an Ashkenazi Jew your ancestors may have spent a 99th percentile amount of time indoors and picked up some adaptation, if there ever was any effect of CO2 per se. In hunter-gatherer and farming times maybe high blood CO2 sort of signaled "I've been exercising too much and need to rest" but after urbanization it meant nothing.
As a practical matter going for a walk outside in the fresh air feels good and elevates my mood. I don't know what the active ingredient is.
When you said high CO2 levels could affect focus and decision making, I thought that made a lot of sense - in the context of continuous high CO2 levels. e.g. if you work in a closed-up office but go out at intervals for 'fresh air', you might feel refreshed and able to continue working, but if you stay indoors all day you might feel much worse.
Your experiment seemed closer to the scenario of sometimes going out for fresh air - I'd be much more curious to see an experiment with one group working in the same high-CO2 environment every day, and another in a better-ventilated one.
Watch out for CO2 meters that "calibrate" their sensor by assuming that the lowest level in any 7 day period is outdoor air. This is a terrible assumption, and gives entirely bogus readings that change over time, but that's how a lot of them work.
Instead, you want to get one with dual IR sensors, one of which is used infrequently in order to calibrate the other. (The theory here is that the main source of decalibration is the IR sensor's LED dimming with repeated use.)
Scott’s link is single sensor and is autocalibrated by week. So each week take it outside or use manual calibration mode https://cdn.shopify.com/s/files/1/0019/5952/files/Manual_AZ_0004_Revised_1621.pdf https://cdn.shopify.com/s/files/1/0019/5952/files/TIM10_Desktop_CO2_Temp_RH_CO2Meter_DataSheet_01_8_2021.pdf
One way the competing factors could cancel despite large effect sizes is if your inclination to play is or is affected by a threshold effect. Like a more complicated and convoluted version of measuring the temperature in a room when the thermostat of a heater kicks in, and then concluding that opening the window doesn't have an effect on the air temperature.
Whether
You best hope there is no cognitive problems caused by high ambient CO2 levels. The US Navy has several Boomers on station, underwater, with very high CO2 levels at all times. Any one of them could kill a couple of hundred million people, and trigger Armageddon.
I firmly hope, and fervently pray that our brave sailors manning those vessels are not cognitively impaired.
Please Lord, protect them and us.
Indoor air quality is typically significantly worse than outdoor air (fire season aside). Seems like a plausible explanation here is that CO2 is a proxy for how well ventilated a space is, but that it doesn't have a direct effect on cognition by itself. The original study may have been picking up some other pollutant where Scott's didn't.
Nice self-experimentation. In my personal experience, air and weather changes (like changes in
pressure) affect some more than others. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3996041/
Personal anecdote - I was taking a 3 hour exam in a lecture hall with ~100 other people with no circulating air. I felt like falling asleep due to how stuffy the air was, but others seemed like they were doing fine.
I didn't collect a lot of data, but I used to gauge how well I was feeling mentally by how long it took to complete an easy websudoku.
Might I recommend Tetris or 1010! for visual/spatial reasoning?
My family used to play this all the time as the game Boggle. Fantastic game, I crush all my friends mercilessly if they are foolish enough to challenge me, who has been forged in the fires of family game night for years.
So, as a submariner, I think I should point out that CO2 levels are very tightly controlled with atmosphere control equipment on board. It’s something that the OOD reports to the Captain every watch, and there is monitoring equipment in every space. Nonetheless, there are certainly higher levels of CO2 and lower levels of O2 on board than there are outside the boat - and every sailor on board can tell that it effects their mental abilities. It’s hard to describe, but the feeling when you come up and equalize atmospheres with the outside world and how that effects your mind is very noticeable. Now we certainly do very hard work very well, like you said, but mental effects from lower CO2 are real. Nonetheless, that’s obviously both anecdotal and referring to much greater changes in CO2 level than you are talking about, so it may not be a useful data point.
Can you tell us what levels are typically like?
I found a LUL, and I must tell the world about it.
Scott, I pay you too much to be banned for a innocuous meme. :|
_______________
I hope more people did these kind of highly controlled sniff tests. I love it.
This just seems a priori kind of unlikely. 445ppm CO2 is 0.34 mmHg, and even 3000ppm is only 2.3 mmHg. Meanwhile, as far as I know pCO2 in tissues is around 40 mmHg, so even the highest concentration of CO2 you tried is 1/20 the normal concentration in tissues.
Amazon not only sells CO2 monitors, but also CO2 generators. They seem to work on the principle of adding citric acid to sodium bicarbonate. Of course, any acid, like the acetic acid in vinegar, should work. There must be some ppm of CO2 which has an effect on cognition. This would be approaching the problem from the other end. With appropriate caution, of course.
Also, increased partial pressure of CO2 should drive an increased respiratory rate to clear the CO2 by the usual physiological mechanisms of negative feedback. Measurements should show that respiratory rate correlates well ppm CO2. And at some ppm of CO2 the compensatory respiratory rate should be overcome.
Additionally, CO2 accumulation in the blood lowers the pH, (acidosis), which probably mediates the effects on concentration, headaches, etc.,
Actually, you want to be a little careful with that. When CO2 levels rise high enough to be unquestionably dangerous, they are dangerous very fast, e.g.:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5380556/
"CO2 levels of more than 30% act rapidly leading to loss of consciousness in seconds. This would explain why victims of accidental intoxications often do not act to resolve the situation (open a door, etc.)"
It would definitely take work to get CO2 levels that high, but it could be done in confined spaces if you were fairly careless. That article mentions that an OSHA study in 2015 estimated about 90 deaths per year in the US by CO2 asphyxiation in confined spaces, with a remarkable 2/3 of those deaths being rescuers.
Nice post! Does anything else about air quality affect cognition? I would imagine at the very least it's harder to get work done if it's 50 or 90 degrees Fahrenheit in a room, and probably hard if the air is forest fire level. I often do feel that I think better outside, but it may just be that I think better standing up or walking. Indeed, at least stereotypically pacing back and forth is supposed to be good for pondering things and I feel there is some truth in this; it matches my own experience (not sure of the mechanism). Likewise changing environments may lead to coming up with new ideas. I've had experiences where I was stuck on something and then as soon as I've walked away a new idea comes up (lots of mathematicians do too). Still I don't know if this is about cognition rather than some sort of change in something. Of course, my feelings also may not reflect reality, but I think there's some truth to the idea that changing the environment leads to new ideas.
This might be harder to test. Especially since coming up with new ideas is very different from finding words. But I would imagine most SSC posts don't come to you while you're sitting down at your desks and while you are walking around. And also you find it hard to focus when there is someone else in the room (you said this in a previous SSC post at some point).
Sorry if this became a little incoherent but the general point is that at least my feeling is that there are some environmental factors that do not really affect cognition, and I'm curious which one it is. My suspicion is that there is something to the fact that a change of environment can be good for cognition, which is why so many scientists like walking. I also would guess that this also lends plausibility to the CO2 thing; it's a proposed mechanism for this because they get out of the high CO2-office. It's worth noting whether this effect is real in the first place, and what causes it anyway, if not the CO2.
Well think about where our ancestors spent most of their nights. Tiny tent/igloo/hut with all family members present. Longhouse with an entire tribe. Most of the year, all of the above had a fire lit and minimal ventilation in order to preserve heat. Imagine how high the CO2 levels were, and how low the O2 levels. Yet the ancestors of whoever is alive today thrived in that environment.
High-CO2-impares-cognition hypothesis makes a lot of sense to me, so here are two ways to save it:
1) Your performance in WordTwist is bounded in the type of intelligence CO2 is affecting: say, you play as good as you possibly can at 105 IQ, and CO2 causes your IQ to fluctuate 130 to 110 IQ. Variance in score is explained by other factors - experience, motivation, attention, and so on. Seems very plausible - the game is not that conceptually hard.
2) CO2 affects intelligence in part through motivation (I know I don't want to do anything hard if I can barely breathe!). So then you were feeling very bad (from high CO2 and other factors), you wouldn't want to play at all, and when you were feeling smart, energized and motivated (from low CO2 and other factors), you would do something more productive. If motivation and ability is correlated, and you only want to play at certain motivation, your score stays constant even if your ability is changing.
Hey Scott, could you post scatter plots of all your data here, correlation numbers are really not very informative at all - you can have low correlation despite a very obvious effect, and high correlation when the data clearly looks like a random cloud.
Scott posted the raw data in the comments a bit earlier: http://slatestarcodex.com/Stuff/CO2data.xlsx
Here's what it looks like: https://thelo.ca/co2.png
I bought a MH-Z19B co2 sensor off ebay a while back to wire up to a pi/arduino and save some $ from buying a whole meter, but not too shockingly never got to wiring it up. This might be the inspiration I needed for it!
If I do that and figure out a testing scheme, I'll make sure to come back and post a prereg of some sort.
Scott, you posted this essay within three hours of someone mentioning the Healthygamer Youtube / Twitch channel (on video game addiction) in the current Open Thread. I think you might be kabbalistically required to give it a look.
Link to the comment: https://astralcodexten.substack.com/p/open-thread-184/comments#comment-2572226
I will replicate this for 1 min chess games if somebody pays for the CO2 meter.
5-10 times 3 minutes might not be long enough duration. Do you have data available?
In all levels of school and university, the high CO2 levels causing impaired capability to participate was considered absolutely obvious consequence of spending prolonged time in poorly ventilated classrooms and lecture halls, which are many and numerous in Finland. I don't know if we ever become stupider, but after 1.5 hours, bad quality of air is obvious and people would start feeling sleepy. This was commonly attributed to CO2 from bad ventilation.
Abstract of the article by Zhang et al 2016 you linked to suggests that the effects are not due to CO2 but bioeffluents.
And no, it is not about the boring lectures. I attended meetings of SF club that has poorly ventilated rooms in the cellar of student housing. All activities are fun. Too many people too long time, many people claim it is noticeable.
However, here is another stupid question. Is the impairment effect about amount of CO2 in air or amount of oxygen in the air?
Air contains both, but near 80% of it is nitrogen. In closed space, oxygen is breathed in and CO2 is breathed out, the amount of nitrogen remains the same. To my knowledge, it is inert and uninteresting thing for all relevant parts the human biology. I don't have access to Zhang et al 2016, but in their experiment where they did not find CO2 effect, they added CO2 artificially into air that otherwise circulated freely. I wonder if introduced CO2 displaced nitrogen and oxygen at equal rates, that is, much more nitrogen than oxygen, but amount of oxygen remained much higher compared to enclosed space where humans inside convert oxygen to CO2.
"...many people claim it is noticeable" it = bad air
"I wonder if introduced CO2 displaced nitrogen and oxygen at equal rates, that is, much more nitrogen than oxygen, but amount of oxygen remained much higher compared to enclosed space where humans inside convert oxygen to CO2."
Thinking about this more, it wouldn't necessarily explain the lack of effect in Scott's experiment. Plausible more people in the room -> both higher CO2 level and less oxygen.
We're talking about CO2 levels of up to about 3000 ppm at high concentrations. That's 0.3% of the air. Atmospheric O2 levels are about 20.9 % at sea level, and seem to decline to about 20.1% at 1000 ft elevation. So I would be very surprised if displacement of oxygen by CO2 were a significant factor in this common experience of "bad air".
thanks for pointing the obvious, with great skill I managed not to think about the definition of ppm.
I can absolutely confirm KPier's observations about chess (I play blitz games at chess.com) - there are clearly good days and bad days, they are mostly unpredictable (ie. unrelated to sleep deprivation etc.) and the difference between them is unexpectedly large
Registering a prediction before I read your results: it'll be one of those maddening things where there's an effect but it's not significant and it's much smaller than the previous studies, so you decline to draw a conclusion other than "so, maybe?". I'm four paragraphs in and the admission that you're playing the game because you wanna rather than for the experiment, while reasonable as a life choice, smells like a potentially strong confounder and I had to pause to think about why.
I believe what I'm anticipating based on own gameplay habits is a self-selection effect where the more up to cognitive tasks you're feeling, the more you play the game. Then if CO2 is only one is several factors in your congnitive state, even if it's the largest one, I would expect no correlation at all - your high-CO2 gaming will occur when other factors combine to partly, if not totally, cancel it out.
One thing to check is if you're actually measuring your CO2 level at all. Most of these cheap meters just auto-calibrate such that whatever the lowest physical concentration they see over a 24 hour period is set to a given internal concentration, usually around 300 ppm.
That said, you'd probably still see large swings. Probably.
I just wanted to express my admiration for the fact that you wanted to replicate this study on your own, and went to such lengths to do it. This is a good win for the scientific method.
I think for certain intellectual activities, there's an optimal level of intelligence. If you're more intelligent then that, you'll want to do something else. Magnus Carlsen made this point about chess[1]. I assume it applies to the wordgame you play as well. This would mean that when you had a low CO2 concentration in your room, but still wanted to play the game, there must have been something else negatively affecting your intelligence at that moment, otherwise you wouldn't have wanted to play at all.
[1]: https://en.chessbase.com/post/magnus-carlsen-on-his-che-career
Could level scaling in the game confound any attempt to find (any) correlation (at all)?
I don't know this game, so this might not be applicable at all. Just throwing it out there.
But I'm guessing you would play many rounds in a single sitting, presumably at the same level of CO2. If you do well/not-well in one round, will the game compensate by giving you a harder/easier match next round? I'm sure there are game designers in the audience who can explain better than me why this is good/bad game design, but either way, many games do something to this.
One way to try to see this could be to use only the first game in a sitting as data. I don't have the skills or time to do this properly from here unfortunately.
Two related pet peeves of mine:
1) This worked for me so it is universally applicable.
2) Studies say this doesn't work for most people so it won't work for me.
Thank you for not falling for either of them.
Hypothesis for confounder: your brain switches signalling strategies between ones that are more or less CO2-tolerant in a predictive way. It uses your circadian rhythm, metabolic rate, etc. as independent variables, and observed CO2 as the dependent variable for this prediction. Your brain manages to function fine (at least in the short term) in both “modes.”
Then, it would only be when your brain mispredicts CO2 and stays in the CO2-intolerant mode while actually experiencing high CO2 load, that you’d see the conditions from these studies.
Under this hypothesis, your brain would probably re-predict easily enough when going from outside to inside, or when opening/shutting windows; so I’d guess you’d have to do something more like “releasing a bunch of CO2 into the room without otherwise changing AQ” to trigger the effect. (Which, if I understand correctly, is similar to what the studies do.)
Geez! I would have blown off studies like this as "possibly bullshit, but do I care enough to try to confirm it? Nah."
But you go out and science the heck out of it! I'm awed.
This makes way too much of a very flimsy and poorly conceived informal experiment, if I understand it correctly. Start with an environment where CO2 levels fluctuate quite a bit (from about 500 ppm to about 3200 ppm), measure CO2 level just after playing WordTwist. Calculate whether there’s any correlation between CO2 level at end of game and score.
In other words, the only thing that was tested is, correlation between momentary exposure to CO2 to game score. No attempt to measure effects of continuous prolonged exposure is mentioned. Yet results are used to challenge more formal experiments that are measuring prolonged CO2 exposure, such as exposure of 8 hours.
This WordTwist experiment is of no significance, unless you hypothesize that the significant effects of exposure to high levels of CO2 is immediate and very transitory. I don’t think there’s any good reason to think that, and it’s certainly not what the studies that are being challenged presume.
It’s an important health and social issue as governments have been supporting or requiring, as a conservation measure, tighter construction to seal in heat, as well as indoor pollutants like CO2, which can easily build up, for example, overnight in a tightly sealed bedroom. The concern is, effects after such a prolonged exposure. If you want to do a realistic informal experiment, try, spending 8 hours in a room where CO2 builds up to say 2000 ppm or higher the whole time. (Well, don’t actually try this by hanging out in a poorly ventilated room full of people until after the pandemic, please.) That would be equivalent to the studies that are in question here. I’d even be willing to bet, your WordTwist scores would show a downward trend after several hours in this scenario.
The posting also asserts that cognitive impairment due to CO2 buildup in the ranges we’re discussion (like, 1500 - 3000 ppm) isn’t a concern on submarines, for example. There’s a comment below from a veteran Navy submariner who reports cognitive issues from CO2 buildup was indeed a big concern in practice. Though, as you note, there’s a Navy study claiming to have found no cognitive effect from even much higher levels of CO2 exposure — but the Navy only tested cognitive effects after exposure to high CO2 levels for 45 minutes, that is, intermittent exposure. Might make sense to test that if as the comment below suggests, steps are already in place to prevent longer term crew exposure to excessive CO2.
If you’re worried about whether a momentary variation in CO2 exposure is going to fry your brain, this WordTwist test should be reassuring, though I’d wonder why you’re worried about that. If you’re worried about a real problem — how conservation efforts may create indoor buildup of CO2 resulting in longer term exposure, this WordTwist test is a straw man.
I moved to a new house and bought 5 CO2 and particle sensors (nerd fact: mh-z19 and sds11 with esphome, 25$/each) to track different rooms. All are recording readings every 10 mins (nerd fact: into my homeassistant and then influx for longterm storage). I also got many cheap (10$/pc) humidity/pressure/temp sensors doing the same and I plan to monitor this for years for the whole 6 people household. Can't wait to see how it (doesn't) correlate with e.g. number of tasks etc.
I did something like this once. I wanted to try clonazepam to see if it would help with some work issues I was having, but I was worried about the cognitive side effects. I coded up the reverse-digit-span test from the Wechsler test as a quick Python program and took it regularly at various times for a week (it takes about two minutes), then started the clonazepam.
What I found was that clonazepam at the dose I was taking had no detectable effect, but there was a huge, massive, consistent improvement in my reverse digit span from taking a half-hour nap in the afternoon.
The thing I have the biggest trouble with is not short-term memory but executive function. I keep meaning to code up the trail-making A and B tests (the reverse digit span thing is a couple hundred lines of code) but I haven't gotten around to it yet.
I've started playing some chess and I also log the CO2 and other stuff, so I can look at the data in some weeks/months.
isn't the way you happen to be breathing during each game a significant confounder here?
I recommend reading "Breath" by James Nestor for some insight into the variability and health effects of different breathing patterns. maybe using something like The Breathing App before each game in the experiment would help?
I am willing to do this and put in a minor amount of effort into a cognitive test (<2 minutes) once when I wake up and once when I fall asleep for about a month if someone is willing to pay for the monitor. Someone will also need to tell me which cognitive test to take.
poast time series graphs
By coincidence, I recently did something very similar (but did not record my results, so everything that follows is impression based.) I played several hundred games of a very similar word game on a 4x4 grid [Boggle with Friends], and found that I performed markedly worse under some pretty predictable circumstances:
- Before coffee
- Sleep deprived
- Drunk
- Drowsy antihistamines (cetirazine)
The interesting part is that board composition (availability of high-scoring words and the supply of T's, E's, S's, R's etc.) had a much higher impact on the score than any of those factors, and that makes me think that boggle-like word games might not be a good metric for measuring variations in cognition.
Also interesting was the fact that the following conditions didn't have a pronounced enough impact for me to remark on them at the time:
- Low to moderate alcohol intake
- Environmental distractions (not touch based)
- Non-drowsy antihistamines (fexofenadine)
- Hunger
- Time of day
And that the following DID have a pronounced enough impact to note at the time:
- Sitting position
- Touch-based environmental distractions
- Attractiveness of opponent (?!)
- Perception of time pressure*
*Notable because the game clock is always 90 seconds, no matter whether or not you feel like you're jamming the session in between two other tasks.
I'd be curious to also see the effect of other indoor pollutants like PM25 and Formaldehyde.
Don't know if this is a "really good experiment" but I'm interested in testing memory, logic, and creative thinking in the form of solving sunday nyt crossword puzzles (i recently bought a book of 500 of them) in a number of fixed environments (eg desk at home, sofa at home, outdoor patio, coffeeshop, park)
I am interested in tracking co2, particulate matter, temperature, humidity, light (brightness, natural vs artificial), visual stimuli (eg plants vs people vs nothing biological, clean vs cluttered, movement vs still), sleep quality/quantity, diet/weight, and time of day (in addition to solving "performance" obviously).
The meters required for this experiment are listed at: https://www.amazon.com/hz/wishlist/ls/1P972IIAIEZZI?ref_=wl_share
If the meters in the amazon list are purchased for me, I commit to solving at least 1 puzzle a day until the book is completed, share the results with Scott (and anyone who's interested), and to ship the equipment to whomever Scott directs me to upon completion of the study (I'm a minimalist and object more to there being more "junk" in my life/the world than the actual cost of the equipment).
Addendum: I would also track subjective physical/exercise activity, altitude and blood oxygenation (I own a pulse oximeter and its low utility makes me highly resistant towards buying the other gear).
I intend to travel quite a bit on extended road/camping/rv trips over the "study period" and think the data could be pretty interesting.
If 3200 ppm causes headaches and nausea, consider me surprised that it has no effect on cognition. Did you plot your data to see if there's an effect in the highest range?
Quite simply the reason you failed to notice any difference in performance related to CO2 level is because you are not a plant.
The experts say the Earth is warming due to increased CO2 levels caused by humans and the graphs show a bump in the recent past although, nevertheless, we are technically still in an ice age, which has been going on for 2.6 million years, with 100,000 year ups and downs.
Currently there's a human caused CO2 increase. Sixty years ago CO2 was only 380 ppm, whereas now it is 413.93 ppm. When CO2 increases, plants thrive and the Earth greens.
Plants inside a greenhouse prefer a ppm of 900, which boosts plant growth and health from 25% to 100% depending on the plant. As for humans the current 413 ppm is no problem. The CDC says you need 100,000 ppm of CO2 to die and other sources say 40,000 ppm is unhealthy for 8 hours. A submarine's scrubber turns on at 8,000 ppm.
Earth has had much higher CO2 levels than the current 413 ppm and our ancestors apparently thrived. 150 million years ago it was at 3,000 ppm and 6,000 ppm farther back. The cyclical CO2 increase is caused by the oceans' warming which releases a bit of the 93% they contain.
Current climate science may be putting the cart before the horse and one might question the causal direction that 97% of professional climate scientists agree on. But experts are purveyors of groupthink - it's how they make their money and continue their funding; they should not be trusted to give a definitive answer.
In the 1980's and in 2005 Phillip Tetlock studied experts' predictions and found they were no better connecting with reality than anyone else. One might spin this result by saying that we live in a random universe and the experts correctly identified and properly proved that fact.
Thus, regarding CO2, there's a 50% chance we really need to produce more of it. Carbon credits should be carbon debits and we must increase our individual carbon footprint to save the planet. But we could be wrong. To best determine the correct answer what's needed is a monkey throwing darts.