Bayesian Truth Serum // Using MATH to catch liars

Get better at math with Brilliant ► brilliant.org/... Learn math more effectively for free and the first 200 subscribers get 20% off an annual premium subscription. Thank you to Brilliant for sponsoring this video on the Bayesian Truth Serum
Check out my MATH MERCH line in collaboration with Beautiful Equations
►www.beautifule...
How can we figure out how many people cheat on their homeworks or their taxes? We could just ask them, but they might lie! In this video we will explore a fascinating piece of mathematics called the Bayesian Truth Serum that leverages the power of Bayes' Theorem and Bayesian Analysis to create a special type of survey that incentivizes telling the truth. Telling the truth actually becomes something called a Bayesian Nash Equilibrium. Using this technique, researchers were able to find a much higher prevalence of contract-cheating on homework than previously reported.
COURSE PLAYLISTS:
►DISCRETE MATH: • Discrete Math (Full Co...
►LINEAR ALGEBRA: • Linear Algebra (Full C...
►CALCULUS I: • Calculus I (Limits, De...
► CALCULUS II: • Calculus II (Integrati...
►MULTIVARIABLE CALCULUS (Calc III): • Calculus III: Multivar...
►VECTOR CALCULUS (Calc IV) • Calculus IV: Vector Ca...
►DIFFERENTIAL EQUATIONS: • Ordinary Differential ...
►LAPLACE TRANSFORM: • Laplace Transforms and...
►GAME THEORY: • Game Theory
OTHER PLAYLISTS:
► Learning Math Series
• 5 Tips To Make Math Pr...
►Cool Math Series:
• Cool Math Series
BECOME A MEMBER:
►Join: / @drtrefor
MATH BOOKS & MERCH I LOVE:
► My Amazon Affiliate Shop: www.amazon.com...
SOCIALS:
►Twitter (math based): / treforbazett
►Instagram (photography based): / treforphotography

[@thoperSought]

re. "contract cheating": teaching English in Japan, I've had two students I know for sure use(d) machine translation to do their homework. I'm not working in a school or college; there are no grades or tests-these people pay me to teach them how to speak English because they want to learn to speak English. this makes no sense to me, at all.

[@chessplatypus4769]

parents forcing them to do it? asuming their adults i have no clue unless they just are to lazy to do homework

[@stephenchurch1784]

Massive cultural anxiety around failure seems like a plausible explanation. When I was in japan there were police in schools and dorms around exam time because suicide rates skyrocketed. Even if grade pressure isn't there, the feeling that not doing perfectly on any one task dooms you to failure forever can be hard to shake

[@thoperSought]

@@chessplatypus4769 yeah, one's middle-aged, and the other is retired.

[@thoperSought]

@@stephenchurch1784 not impossible; there's not really anything to fail at, though. I think for the current one, who's (semi-)retired, that he does it because he really wants to talk about big important (to him) topics, and he wouldn't be able to approach those topics without writing it in Japanese first and then using automated translation. one problem is that the way he writes in Japanese is extra hard for the software to translate-he never uses an easy, clear word when a difficult, ambiguous one will do, for example-so it's really hard for me to understand what he means a lot of the time.

[@4.0.4]

@@thoperSought Well, have you told them you know they're using machine translation (and how bad it is)? How did they react?

[@Lucassis]

Q1: Did you pay someone to complete your homework? A: No (truth) Q2: What percentage do you think have paid someone to do their homework? A: 95% (also truth) Q3: How can this be? A: In a class of 20 students, I was the one getting paid.

[@Emily-fm7pt]

Negl I actually have gotten paid to do homework for classes that I wasn't even in... Should I feel bad, probably, but do I feel bad, no... (It was for a computer science class that some med students were taking, and they just couldn't grasp the concept of a loop even after spending a couple hours trying to explain)

[@yusufklc7821]

This is me in pandemic hahaha

[@livedandletdie]

@@Emily-fm7pt Wait they didn't grasp the concept of loops... Like while and for loops? How the hell did they pass kindergarten? How much of an NPC can one be. They all have 2 brain cells fighting for 3rd place.

[@cortster12]

This really would not work well for pessimistic people, lol. They'd say no, but then think so poorly of their peers it's like 99%.

[@legendgames128]

"I do not cheat, because over 101% of my peers cheat, and I do not want to be like them."

[@bethanywaanders8579]

There is a difference of sensitivity and interpretations of actions too. With lot of grey area around collaboration vs copying, tutoring vs homework help vs contract cheating it’s blurry but reliable that a wide majority of people overestimate themselves self identifying as above average (good, honest, smart, kind, whatever) when that’s highly unlikely and many are below average.

[@buckeye998]

I think it is more generally for a population. It won't prove 1 person is lying, but it can be a pretty good rough approximation for what percentage of students are cheating.

[@troublemack9876]

That is why they take an average over a large sample. They try to mitigate the impact of outliers

[@sr2291]

People cheat on tests by glancing at your paper. One of my professors used to give tests in a large room and naked everyone separate two desks away from each other. You should have been there to hear the grumbling. Lol

[@johnchessant3012]

I thought this was going to be the trick where the survey has you flipping a coin, where if it's heads you always answer "yes you cheated" and if it's tails you answer truthfully. But this is even more interesting!

[@benwichser2991]

I was hoping the video (or a follow up?) shows how this Australian method is better than the coin-flip method. But it was fascinating nonetheless.

[@cagedgandalf3472]

"I am a student cheating everyday, I now know how to answer the surveys. Thank you."

[@shacktoms2791]

It seems that the implication is that "telling the truth" usually means "confessing". If people think they are being rewarded to answer as though they are cheating, then how many will do that even if they are not cheating?

[@tomasrydholm9144]

Maybe I'm missing the point but wouldn't this encourage the students to contribute as much as possible to a high iscore, i.e. by ticking "Yes, I cheated" and guessing that 0% cheated? I do see that people who are in fact cheating has something to win on telling the truth, but doesn't it make people who didn't cheat more likely to lie?

[@stephenchurch1784]

I've got nothing to back this up but I have a feeling that there is some correlation between academic integrity and propensity to lie on a survey. I doubt it's a one-to-one correlation but the situations are both tests of integrity

[@IsomerMashups]

@@stephenchurch1784 The motivation to cheat on schoolwork is different from the motivation to cheat on a survey.

[@ajreukgjdi94]

I thought the exact same thing, but I realized, the iscore for a yes answer would increase, but the frequency of ticking no and the probability guessed of people ticking no can also be used to create an iscore. This iscore would drop by doing this

[@seanmccloskey3816]

Not clear but maybe your iscore is based on your answer on whether you personally cheated. I.e., if you say you did not cheat, iscore measures positive surprise of percentage who didn’t admit cheating vs percent you expected to not cheat. One thing he doesn’t address is survey size. If there is a large survey, your confession is going to have a minimal impact on the iscore.The whole Bayesian survey design might still have predictive value because of the effect he describes of believing most people are like you, but large survey size will dilute the incentive to be truthful.

[@samprentice9454]

if you underestimate, i thought your iscore would be lower?

[@BirilliantSkyStar]

What if a genuine person thinks many other people would cheat on the same difficult exam? Isn't this survey biased to only assume that genuine/not genuine people would think other being genuine/not genuine?

[@nalat1suket4nk0]

I was thinking the same thing, if someone actually didn't cheat, nothing stops them fir thinking other people cheated

[@MrWizardjr9]

or maybe they dont cheat but they know some people that cheat which will still psychologically bias them towards thinking more people do it

[@TheRationalPi]

I remember a take-home exam in college where nearly two-thirds of the students in the class worked together on it, despite *signing their name on the first page of the test to say that they would not use outside help*. They got caught (prof said they all used some weird notation that wasn't taught in class), and each received 0 points on an exam worth a solid 20% of their final grade. The heuristic in this video would probably mis-categorize me, because I've never cheated but I'm now convinced that pretty much everyone else regularly does.

[@VietnamSteven]

I can't believe that Bayes could go from Medical testing to Cheat detecting. Incredible! Thank you for providing this brilliant video!

[@satiatedpanda]

Really great video! One question that might want to be added is if that person *knows* if another person cheats. If a person who doesn't cheat, but knows another person cheated, will assume that most people cheated and be marked a cheater in this test.

[@Lantalia]

This method is going to pick up a bunch of neurodivergent people that are answering honestly, but angry about others behavior

[@lordhater4207]

OMG! He figured out the thing that psychologists seem to be unable to : "Sometimes people lie".

[@idontwantahandlethough]

you think psychologists don't know and account for that...? That's like the first rule of the social sciences.

[@lordhater4207]

@@idontwantahandlethough Oh i know ways, all their conventional ways are ineffective and the only types of research that would yield somewhat useable data is unethical and would not pass ethics committee, best they can do is what they are doing using flimsily barely proven concepts and pile their barely provable hypothesis on top of those flimsily barely proven concepts. Even Bayesian Truth Serum has gaping holes, you cant statistically account for maliciousness because malicious people are by their nature unpredictable, all you can do is construct a rigid narrative and compare everything to it but by that point you are no longer dealing with science but with religion or a cult, which is where i see psychology at, although i have to say honor to the exceptions however there are few of those.

[@blahybris608]

@@lordhater4207 most sciences have their unknowables. It's like saying physics wasn't a science before we observed light's properties as a wave and a particle or before we proved quarks and stuff or before we figured out gravity (oh wait, we didn't really yet, i guess physics must not be a science)

[@lordhater4207]

​@@blahybris608 Yea i know psychologists like to compare themselves to physicists or mathematicians however that's like comparing apples with anvils, physics is a WELL very WELL defined science because it's research is limited to physical realm be it infinitesimally small or incomprehensibly large it's still within physical realm, physics can be wrong about their hypothesis however if a hypothesis is wrong only that hypothesis is wrong and perhaps its branch. I like to use reverse engineering as an example, physicists can see memory of the system (nature) and corresponding values of STATIC memory addresses point (both discovered and undiscovered laws in physics) at they assume how background processes (that which they can't see or understand by observation alone) produce values they see those STATIC address point at ,and test it out, when and if they are wrong it's not the fault with STATIC address but of their assumption, they can backtrack what they did wrong since subject of their study is static and so they can simply pick themselves up and start over having learned even in failure since their fail may just be data type and not wrong assumption, however physicist cannot CHANGE variables because if that were true then we could turn gravity off, physicist is not in control of nature it's just its scribe it's documentor and observer, this isn't true when it comes to psychology psychologists may see memory of the system (human) however their problems are way more complex than the one physicists have all due to a fact that their memory addresses aren't static they are dynamic as in they change due to all sorts of reasons from subject of their study (humans themselves) to the methodology used to uncover background processes (research) that when they are wrong and they often are cannot even bactrack what went wrong because by the time they do address had changed and they don't know if it is environment, is it the subject , is it methodology in other words psychologists due to the research can and do change variables and can influence outcomes, this doesn't happen in mathematics or physics and since it's simpler when compared to psychology they can do some really impressive stuff with it some highly complex experiments and still be valid because subject they study is static, whereas what psychology would do with complex experiments is only introduce yet another point of failure, thus is the reason why they limit themselves on questionnaire "research".

[@leop5180]

@@lordhater4207 chill bro

[@readjordan2257]

In my school the professors say im supposed to just copy from those that know...and use that as the learning process. Apparently, homework is supposed to always be a social event. Wheras exams are not.

[@Polymeron]

I'd need to read the original article and the reviewers' comments, but off the top of my head, if you financially incentivise people to give surprising answers, you will get surprising answers. Suppose I didn't cheat, but I want to maximize my iscore so as to get as much money as possible from the experiment. I could answer that I do cheat, thinking that as a rare thing that people will predict to be rare, my answering this will nudge up the iscore for it. More, I could guess that others might have the same rationale, boosting it further! This, whereas on a survey with no incentives I would likely answer truthfully that I didn't cheat. This creates a biasing of the responses. So great, they get a result with a much higher percentage of cheaters than before - but how do they know those results are more truthful than the standard survey? Again, I'd need to read the paper, but I'd argue that they can't know that.

[@tudornaconecinii3609]

"More, I could guess that others might have the same rationale, boosting it further!" Guessing that most people have this level of meta-thinking would be irrational. You can look at past studies of this kind and see how many people act GT-optimal or adjacent, and how many just do what seems prima facie correct. It's actually ironic, because you are engaging in literally the same bias relevant to the bayesian truth serum. It's an example of "You are more likely to believe X is common, if you do it yourself." but with the variable X being equal to "thinking at multiple metalevels" rather than equal to "cheating".

[@Polymeron]

@@tudornaconecinii3609 I didn't say "most" - I said "some". Assuming I would be the only one to think of it would be irrational in itself. Either way, whether I am overestimating the portion of people who would game the system or not, so long as it is not zero this would still produce a bias - rather than produce more truthful results. Glancing at the research, it seems they did try to control for such effects, so I'm not that far off the mark. However, I haven't yet read the whole thing.

[@tudornaconecinii3609]

@@Polymeron Your overestimation of the number of people gaming the system is directly relevant to your instrumental goals here, since it skews your answer to the second question, which results in you getting less money. I'm not proposing that you assume you'd be the only one to think it. I'm proposing precisely what I proposed: namely, to look at past studies of this kind and see how many people act GT-optimal or adjacent, and how many just do what seems prima facie correct. Because knowing to what extent people are superficial thinkers will directly boost your accuracy in answering questions of the second kind.

[@Polymeron]

@@tudornaconecinii3609 It still seems irrelevant. First, If my estimation of people's estimation for the question is very low, then just my lying alone should make the surprising result more common than expected. Secondly, my point was specifically that people being encouraged to write surprising results, can introduce a bias - rather than only eliminate an existing bias. You can't both say that people will not act rationally and thus there will not be a biasing effect, and *at the same time* claim that the introduction of the incentive *will* have an effect because it changes people's behaviour as they try to maximize their gains. (Note how neither point hinges on my throwaway comment that an actor thinking of lying may further consider others doing the same as further incentive to lie)

[@tudornaconecinii3609]

@@Polymeron "(Note how neither point hinges on my throwaway comment that an actor thinking of lying may further consider others doing the same as further incentive to lie)" I am aware of that. I am not saying "I think you're wrong on this particular point, which breaks down your overall argument". I am saying "I think you're wrong on this particular point, full stop." I agree with your overall argument. I just didn't think of mentioning that.

[@ErmisSouldatos]

For this to work, the students have to make the calculations described in the video, but if they are cheating to pass the class, they will not have learned enough for them to be able to take all of the available information into account correctly

[@Hi_Brien]

Oh boi, do a whole 10% of people ACTUALLY pay others to do their homework for them? I'd have guessed 1 or 2 percent... people with more money than brains.

[@РоманБурый-х4б]

Yeah, why would you pay someone to do your homework if you can ask your friend for free?

[@Hi_Brien]

@@РоманБурый-х4б hahaha

[@skinnybonesjonestheallseei74]

@@РоманБурый-х4б or just beat up a nerd🤓👆

[@Zenuku]

damn. I would have guessed around 20%... but I dont cheat

[@johnohm8067]

Truth be told I rarely thought that individuals cheated until I talked to a professor and over heard a conversation of students who did cheat. I am appalled that some students do cheat on stuff

[@alexz.t.4172]

I didn't expect that the correct spelling is not "theorem" but "serum"...

[@readjordan2257]

2:01 as someone whos taken statistics courses... That second question is flawed. You have to always put in a neutral or "i dont know" type of option, or else your survey is unreliable.

[@lidoevera]

I just found your channel one week ago... I've been binge watching every single video you uploaded, you have a great energy when explaining things, I love it! Could you share the doi/title of the paper you were analyzing?

[@skydude7682]

My guess was based on two factors who would cheat if given the chance 40% the amount that actually would go for it would greatly depend on perceived difficulty and consequences about 20%

[@Just_Moh_it]

What if ~90% of the class cheated, and the 10% who didn't cheat get a bad score on the survey? How would you identify the group that cheated (good survey scorers) from the bad scorers, or is it an assumption that the majority is the one that doesn't cheat Great video, fascinates me actually. Subscribed right away!

[@DrTrefor]

Thanks for the sub! In that scenario, those who didn't cheat would still be likely to think cheating isn't as common as it really is, so they would still think they should tell the truth to get the biggest score.

[@HarmonicaFag]

I wonder if taking into account another factor: subjective judgement of how harmful particular action is, what would data say. Maybe we could find higher probability that someone is lying if someone’s opinion about harmfulness of a particular action deviated from the norm in one way or another.

[@taleladar]

So all this "data" is based off of people's responses, which may or may not be truthful, and it also factors in some completely wild guesses, where they are asked to pick a percentage out of thin air. On top of all that, lots of psychological bias when designing and interpreting the data, and bribing people to get a good "iscore". Even after that, all you get is a probability. Conclusion: the results are completely dubious at best, and worthless at an individual level.

[@debu6157]

Me who's gonna break the system by putting no i didn't cheat and 100% them cheated

[@wudgee]

In exams there were always those who went to the toilet a few times. Seriously, I don’t believe they needed to go,and were looking at information in private. Although I’ve never cheated my observations would make my estimation of cheaters quite high.

[@twistedsiblings7544]

Trefor is such a goober. Old UC student here. Love you

[@burngrace5205]

I don’t get it

[@l8dawn]

I've been in classes where I know at least a third of my classmates are cheating. I never have, but if I put down 33% that obv would've been seen as me lying about cheating. I really hope this thought experiment is never actually used cuz it's very flawed lol

[@wjrasmussen666]

This isn't proof someone cheated.

[@asagiai4965]

"I can attest that a lot of students don't do their Homeworks" ~A Paid Homework Maker

[@vojtechsejkora1554]

I do not understand, why do not lied that we do not cheat and select that nobady cheat. So I minimaze probability a lot and still can get iscore?

[@astroceleste292]

exactly

[@CrooningRevival365]

Idk man. I never cheated on a test and yet I believed a lot of people around me did. In fact, when I was younger I believed that I was somehow ‘better’ than those around me because I didn’t cheat and it was an active motivation to keep me from cheating. I feel like there’s some further psychology to explore here. Maybe I’m just an outlier. Wait…no that’s what got me here in the first place. 😂

[@polbarrull]

Big retweet over here

[@julianfogel5635]

I paid someone to watch this video for me so that I could say that I saw it myself.

[@makingtechsense126]

I am kind of a pessimist and I think a lot of people cheat despite not being a cheater.

[@MultiUltimater]

You can also add more questions like this but where the answer is more obvious to get a sense of if this person is a higher chance of lying in general than others.

[@telnobynoyator_6183]

I like the following variant : Before the test ask people to think of a number then roll a dice. If the number they choose matches the number on the dice, they have to lie on the question. That way there is less incentive to lie, because if the truth is embarrassing, you can always say that you didn't really write the truth and that you lied because of the random number. This somewhat attenuates the "lying to yourself" problem but not completely. Then when you get the test results, you'll know that about 1/6 of them are false, which allows you to accurately compensate for that.

[@Xonatron]

Interesting!!

[@wachyfanning]

Watching this so I know how to avoid the Bayesian Truth Serum

[@pikminman13]

I think the thing about cheating is that these days school isn’t about learning it’s about the piece of paper at the end and if you end up having to relearn it anyway proving you learned it once means your capable of learning it. I’m not saying it’s a good thing, believe me, it makes me sad. But nobody wants to spend hours and hours of their week doing homework especially for classes that are required for graduation and basically meaningless otherwise.

[@vapourmile]

Damn. I thought you said "Have you ever cheated before on your homework or your cactus". I had to check and was disappointed when I found out you didn't say that. Do you want people to cheat on their cactus? What's wrong with you?

[@byronwatkins2565]

Actual adults are also capable of witnessing other people acting poorly without having a desire to imitate them.

[@oldscooljoe6194]

Wait, homework is worth something in the us? Here in the netherlands there is a lot more freedom homework wise. But we're told not to come crying when we dont do it and get a low grade. Homework is 100% free choise for me.

[@DrTrefor]

I’ve generally found students are must less likely to do homework that isn’t for marks, but will do it if there is the incentive there

[@oldscooljoe6194]

@@DrTrefor hmm, interesting. Well its also true that we here in the netherlands start experimenting with homeworkless education. Which is showing some promise.

[@EminentInception94]

Ain't never heard of no Beijing tooth syrup before so thank you 🎂

[@legendgames128]

Some things worth noting about this test: Pessimists will likely get marked as cheaters, even if they don't cheat, because they will think bad of their peers. Optimists will likely get marked as innocent, even if they do cheat, because they will think good of their peers. Cheaters who know this info will likely have a better chance at being marked innocent.

[@shanz7758]

The thing is, people live for a long time, and if the question is "have you ever cheated", I think it's safe to say that someone who goes on higher degrees studies will have participated in thousands, if not tens of thousands of controls / homework in the course of his life, and the probability that someone has actually never, ever, ever cheated whatsoever, is in my opinion pretty slim, by the simple fact that we are humans, unperfect humans, and even though cheating is bad, cheating once in a lifetime isn't a big deal, and even people that are profoundly calm have surges of wrath sometimes. So I'm not sure overestimating the number of people cheating is always a sign of cheating in on itself. It could just be that you are cynical and unfaithful of humans.

[@varunramakrishnan7676]

Humans are surprisingly consistent

[@maxgoldstein6309]

9:51 Yep, I cheated with some frequency in school and would have guessed around 5% or lower lol

[@maxmuller445]

This assumes noone answers "Yes, I did cheat" if the haven't. But maybe there are people who just want to skew the result in an direction in order to have a guessing advantage over the others?

[@dmh20002]

Thanks, now I’ll know to answer 0% if I get a frequency question.

[@ThatJay283]

Well now I now how to treat bayesian truth serum

[@stevezelaznik5872]

Isn't the incentive for everybody to guess that 99.9% of everybody else is cheating? That would give them the highest iscore. Wouldn't you want to reward the person whose iscore is closest to zero?

[@Low_cops]

this isnt what i wanted when i searched BTS...

[@coder436]

here's a different method: just ask the question and how many weeks ago they learned how to answer it, so if they're cheating then they would probably get the answer wrong

[@Zenuku]

or you may get the answer wrong because you remembered the information and not the exact date you learned the information lol

[@wolfgangengler8088]

Rating popular opinions about perceptions of popular ethics is no different than rating anything else. Science and math around inter-rater and intra-rater reliabilities already has a rich and well-established research field. Look up common IRR statistics, like unweighted and weighted Kappa statistics, whether of Cohen or Fleiss forms, or the rather complicated approach of Krippendorff's Alpha, or the relatively simple converse of Gwet's AC1.

[@matthewfarr1107]

Having earned three degrees I actually had to opportunity to OBSERVE actual cheating in action once, and the result of cheating another time. In the former case, it was absolutely blatant. In the latter it was pretty obvious that someone got a hand because said individual made it into an elite honors-style program and was clearly not bright enough enough to play. The head of school (who is a dear friend to this day) didn't have absolute proof of the cheating but the fact that the person failed to reveal a felony conviction made it easy to deny a licensure recommendation. Aside from that, I am at my core an honest person, and I like to think the best of others until they prove otherwise, so I would fail horribly at winning the money!

[@Yupppi]

And this is how vsauce2 got all the scandal math court cases where they used probabilities to define guilt in an individual case and convicted innocent people based on "well, there's a good enough chance you did it in our opinion so it must mean that rare cases don't exist, you're as good as guilty". I wonder how skewed the conclusions would be if the probabilities data was from another area than the actual test, like say using a different country's data, since I assume the cheating rate to be very culturally related in terms of how acceptable and common it is. For example in my understanding finns are rather honest folks, but I know a lot still cheat either in things that don't pose any meaning to them so it's ok just to get rid of it, or occasionally in things that are really important for them to pass. Low cheating rate would be observed in things where most of the people think of it as the measure of their learning and being honest and true would benefit them the most over passing for free.

[@micayahritchie7158]

5:06 I lik cricket. I find I always significantly underestimate how many people like cricket where I'm from because I know it's a sport of declining popularity. Does this type of logic invalidate the test?

[@OMGclueless]

Isn't there still potential for the sensitive question to bias people's answers to be _too high?_ If I were taking this test and (1) I recognized the sensitive nature of the question and (2) thought that people would under-report the true amount they cheat but still give high estimates to the rates of cheating, then I might want to _lie and report that I have cheated_ even if I haven't, to make up for those people.

[@DavidBeckwitt]

Great video! Thanks!

[@DrTrefor]

Glad you liked it!

[@livedandletdie]

Take my word for it, I've not cheated on taxes, homework, or hit the dislike button on your videos. I live in Sweden, I just have to log in to the website of the Tax agency, and click a few times and then log off, and it's done. I didn't do homework, homework is useless, and does nothing for learning. And the amount of YT videos I've pressed dislike on, is despite it being more than a decade spent on here, significantly less than 100, it is most likely greater than 2 and less than 10, but I'm not certain, it is hard to remember 10 years of activity. And I can't say that there have been a great multitude of videos I've liked either. Also probably less than 100. So in general I am an honest person, however if you hand me a deck of cards, you'll bet your arse I'll be doing my best to stack them in my favor.

[@IsomerMashups]

There are also the people who _know_ a ton of people cheat but don't personally do it.

[@SillyPutty125]

Lots of the people in the comments saying "I think a lot of people cheat, but that shouldn't get me labelled as a cheater." The study doesn't predict whether you are cheating based on how much you think other people chat. It uses statistics to predict the number of people that are lying, but can't tell us who they are.

[@lumen6378]

I ve already lied But i don t remember having ever cheated in my life

[@luisrosado7050]

5:13 sorry for double commenting, though ive internationally made a dramatically high guess in my mind dispite never cheating, mostly cuz i think the idea of cheating is pointless and takes more than it gives But i settled on anywhere between 40% to 70% exclusively, i honestly dont care how many people cheat, im not gonna join that crowd of time wasters when im trying to ACTUALLY get knowledge to PRACTICE my career and follow my dreams, seriously, if you spend hundreds of hours and resources to not learn then just drop out, am i right?

[@lysander3262]

Now do the Iterated Cheating Survey and see what strategies develop! Because capital t Truth is definitely something that is revealed with monetary incentive...right? Alternate punchline Student: "But where do prior probabilities/actual frequencies come from?" Master: "That is the deep Bayesian magic, not for the faint hearted or the monetized video."

[@rollingdigger19]

Did you cheat? No. Cheat did me.

[@ТимофейЧерников-щ2х]

Comparing cheating on the homework and cheating on your partner is like comparing killing a flea and murdering a person

[@luiss428]

i mean for this question to come up theres gotta be some reason why students cheat right? of course you can say that its negligence, but theres a lot more factors going on besides just "not caring about studies" i guess thats off on a tangent. interesting video nonetheless

[@luisrosado7050]

0:08 no, actually this is the first time ive seen one of your videos

[@mlagathomason2555]

The average student paying for people to do a Google search

[@archbox8593]

So doesn't this method depend on the students not knowing about "Bayes Truth Serum"? Because i could imagine a "cheater" ticking "no" and then answering a lower percentage than they actually expect if he knew how this test works 🤔

[@sauravdas6436]

If I was in it for the money, and I think the true answer was going to be higher than the people who admitted it, I would say I cheated (whether or not I actually did)

[@weishanlei8682]

Sounds very interesting! But if you cannot figure out the probability if an individual person lying ir not, what would be the merit of the whole point?

[@DrTrefor]

It is still worth knowing in general how many people are cheating (perhaps incentivizing countermeasures) even if you can't say a specific person cheated

[@autoimmunedefficiencysyndrome]

Homework is BS. Read studies on the efficacy of homework before claiming to give it to your students for "practice".

[@NancyLebovitz]

The people who would know the most about how what proportion pay for cheating would the people who are selling answers.

[@noahanderson8688]

That's also cheating though.

[@aiocafea]

maybe the qurstion should be framed differently is a person selling performance-enhancing drugs 'cheating in sporting events'? maybe you would say yes, i would say no to cheat, i think you have to participate and fix your own results, else you help someone else cheat 'i don't run marathons; i cheat in marathons' sounds nonsensical to me then, you will also have a good or inflated estimate of the amount who cheat, even though you are just part of the 'industry' i see it as 'do you know how many people smoke? do you smoke?' asked to a non-smoking representative of British Tobacco some people will just know the statistics

[@MrBabyBitch666]

This is just a confidence trick. Good cheaters are well aware of these tactics and know how to avoid them. It's like the survey I filled out in HS where they asked about drug use. There was one question that asked frequency of use for a fake drug. Obviously there to nullify results should you say yes to it.

[@lloydgush]

Aren't you then asking "how many people answered yes/no"?

[@vindictiveDOOM]

I don't think anyone cheats

[@uplink-on-yt]

Have you cheated? Yes What do you think the cheating prevalence is? 0% Wait, how? I lied on this survey.

[@Kellythelawyer]

Spiders George said he does not cheat.

[@justins5756]

This is off topic, but I do peoples homework because it is an easy way to make money. I can make more than $15 an hour doing so, and improve my self while doing those task

[@Feds_the_Freds]

Now I wanna know, what we have to answer to get the biggest iscore say, we cheat but think only 0.1% of the whole population cheats? or say, we cheat and think about 90% of the population cheats? or the other way around, saying we don't cheat? Just give me the money :D EDIT: Ok, so I think it's about predicting the answer of the population. Let's say that I think fewer than 50% of people cheat, then I would have to answer that I cheat and think that about 0% of people cheat, to make the answer that I cheat more surprising. Did I win?

[@yourlocalmuslim8789]

Second no cap my drillers

[@DrTrefor]

You must be telling the truth!

[@hw_yozoraVODS]

i've never cheated on homework, but i can see why people do, having to spend 5-9 hours on school, to get home and have to learn more shit i won't use. i love biology, and i hated biology classes, cause for anyone wlse who is not studying to become a biologist is just a jumbled mess, who the fuck cares about the name of the hormone that makes plants seek out the sun? not only that, they could've been teaching how to do CPR, Heimilich, First aid, how to care about your health and the health of other around you. its just so much useless info, that could be replaced with saving someone''s life.

[@King_Of_Games]

I’ll save you some time everybody lies

[@abcq2]

Sorry, I've watched your video three times and I don't understand your explanation. Why does the strategy of "answer truthfully" outperform the strategy "answer with [yes, 0%]"?

[@abcq2]

Your explanation makes it sound like they just score everyone on log(Q1 / Q2)

[@abcq2]

i found the original paper and it says it is trying to leverage the property that "in most situations one should expect that others will underestimate the true frequency of one’s own opinion or personal characteristic" (which makes at least some sense to me.) ==> i really really don't like the use of geometric mean here! estimates of "what proportion of people cheat" are going to be pretty strongly *bimodal* and i would have at least scored people by "what proportion of respondents underestimated the frequency of your response"... but this survey still seems open to abuse. "everyone say yes" still seems like a winning strategy.

[@abcq2]

"proving that a strategy is good" is not as good as actually showing that it's the strategy used by respondents, especially if the majority of the population actually *can't understand why the strategy is good*

[@abcq2]

this is all still only measuring your preference of survey answer, i.e. "answer 'no' if you think people will underestimate how many people answer 'no' on this survey", and people have a preference toward answering 'no' when asked whether they cheat on exams that has little to do with the truthfulness

[@abcq2]

i think this whole idea is horse shit. i think this survey is bunk.

[@FirstLast-ih6ec]

Outsourcing work prepares students for the real world. It allows us to focus on the things we are really good at.

[@onradioactivewaves]

I dont buy this, theres still issues. People could also lie and say they cheated when they did not.

[@lobsypobsy]

What would compell someone to do that?

[@sweis12]

@@lobsypobsy money 💰

[@kylebowles9820]

Lol your outlook on cheating is incredibly naive, it's not about giving up practice LOL it's about lazyness.

[@jimmystrangus2687]

It seems like a gargantuan assumption to say that you will think something is more common if you yourself do it. Is there any evidence of this, or is this just one of those things that "seems" true? I can provide just two counterexamples off the top of my head. 1) I recently finished a Master's Degree in Computer Science. I myself never cheated (you'll have to trust me on that), but just interacting with other students (mostly undergrads to be honest), and noticing that almost every single business unit on the street just across from my University (that would be College Street and the University of Toronto) was occupied by a contract cheating business. So I would estimate the percentage of cheaters to be very, very high. Perhaps you would argue that actual cheaters would score it higher than I would, but that's just an unfounded assumption I feel like. 2) During the "Satanic Panic" that lasted for several decades, certain segments of the population thought that people who listened to Rock N' Roll music or who played Dungeons and Dragons worshipped the devil. I do not think the people who earnestly believed such things to be more likely to be involved in Satan worship than the actual number of Satan worshippers in the population at large, or limited to the rock music or D&D communities.

[@ejrupp9555]

I predict this guy is lying. The number of times IF and IMAGINE were said makes me want to vomit.

[@philotomybaar]

So, I’ve written many papers for peers but have never myself cheated on any assignment. That raises my prediction of cheating frequency and confounds the “serum.” Perhaps it would be better if I were asked if I’d “participated” in contract cheating, to which I’d answer yes. Or, to put it another way- I believe that many people cheat but I don’t cheat myself. So, how should I answer the questions to best earn the money? Should I lie and say that I do cheat in order to get closer to the mean values, or should I use Bayes’s theorem to update my beliefs on how likely people are to answer the first question affirmatively? It’s an interesting question if you consider the problem from the outlier of someone who’s facilitating the cheating but not cheating on their own work.

[@DrTrefor]

A sobering lesson in why the statistic is useful over a large sample, perhaps, but not for any specific individual.

[@mwei2806]

"written many papers for peers" == cheating

[@hedgie9823]

@@mwei2806 No because, they aren't actually cheating in tests, they're just involved in it or helping others with cheating

[@ebob0531]

@@hedgie9823 It is still cheating, most university definitions of cheating include being the provider

[@hedgie9823]

@@ebob0531 Thanks for the clarification

[@SlimThrull]

I'm not sure I buy it. This is based on a cognitive bias. Does everyone have that cognitive bias? If so, is it to the same degree? (I don't have answers to either question but I suspect the answers are: Yes, and almost certainly not). If we can't assume that everyone has the bias or that it isn't to the same degree the numbers become a minefield of best guesses. Perhaps they are accurate, but perhaps they are not. In any case, very interesting video.

[@Yorick257]

Also, the bias depends on the environment. For example, even though I didn't cheat, many of my schoolmates did, or at least were planning to. So, my bias would be in an opposite direction.

[@KingQuetzal]

@@Yorick257 exactly as you say. From viewing my peers throughout HS I would have said 50% cheat. The question would also have to be more specific, what defines cheating in my mind is probably different than other people.

[@Palozon]

Did you finish the video? He tackles this at the end, at 9:38. While it's impossible to identify individual cheaters, the methods here are still useful to make predictions about a population. Variation can and will exist, but in statistics you usually fight variation with large sample sizes and some math.

[@SlimThrull]

@@Palozon Yes, there will be variation. How much? You need to know that in order to deal with how it affects the numbers. Even taking a sample size of a billion would be useless unless you knew what the variation was.

[@JoeARedHawk275]

@@SlimThrull I didn’t watch the video, frankly because I’m more interested in the psychological side than the mathematics of cheating, so I don’t know what method he is using. However, if he is using inferential statistics, then we can use various methods (which you would know about if you’ve taken statistics classes) to find the variation.

[@brandonbolton]

This is the perfect time for this video for me. I am currently learning discrete math and I love seeing real world examples of it! Keep up the great work, Professor!

[@DrTrefor]

Glad you enjoyed! Good luck in discrete math:)

[@bucherregaldomi9084]

whabout discrete maffs doe

[@Drakonus_]

Did I miss something in my Discrete Math class? Cause I actually learned this in Basic Statistics. Maybe I forgot.

[@Celastrous]

This topic isn't discrete math. He's saying he loves seeing examples of real world math because he's studying it.

[@joshua5669]

This is awesome. I am an undergraduate student who participated in the study. One of the writers is my lecturer. It is very interesting to see this explanation!

[@RubyChiang]

Well, if a question such as, “What percentage of people do I think have cheated on a test?” or something like that, I usually answer somewhere over 50% because while I never cheat on any of my tests, I am aware that people do cheat. These usually come from rumors of instances of students paying people to write their essays or buys answers of test questions, etc. Prior to all this, I seriously didn’t think people who pay up to cheat because the consequences were quite severe if you were caught cheating. It scares the crap out of me and so there was no way I thought people will still cheat considering the consequences.

[@jtstearns3373]

I watched pretty much everyone I went to high school with cheat, so I guess I have a notion that its way more common than it is.

[@KingGisInDaHouse]

Do it the scientific way. Have them show you their work on the white board after the class.

[@mimikal7548]

Ngl I don't think this is very accurate since I don't think your assumption that you think something is more prevalent if you do it yourself is necessarily true.

[@meiliyinhua7486]

this is a bit... flawed at least like imagine a classroom of students who mostly don't mind phones going off in class and sort of tune it out, and for whatever reason a no phone rule either doesn't exist or isn't enforced Now imagine handing this test to the one student who is constantly annoyed by phones going off and as a result turns theirs off to limit the annoyance as much as they can