Kachelmeier, S.J., & Shehata, M. (1992). Examining risk preferences under high monetary incentives: Experimental evidence from the People's Republic of China. American Economic Review, 82, 1120-1141.
That's what my notes says was the famous experiment which tested some biases, I forget which, in China using offers of up to several months equivalent salary. Just a fast answer grabbed from my notes, I didn't try to Google or check notes for what the experiment was about.
EDIT: Paper here, seems to be about pricing some simple lotteries. http://decisions.epfl.ch/ExpFinance/readings/KachelmeierShehata.pdf
The GRGRRR problem in Kahneman and Tversky's 1982 paper shows almost identical results whether or not the subjects got a $25 reward for winning.
Well, if someone knows about systematic biases that don't go away with incentivization, they're probably too busy making money off that insight to comment here!
In practice, when the stakes are high, it is not so much that people start thinking more accurately -- though this will happen to some extent, and for some people dramatically so -- but rather that they become more cautious.
If you take ordinary folks into the lab and ask them questions they don't care about, it's easy to get them to commit all sorts of logical errors. However, if you approach them...
Well, I would have done some research and gotten a warm fuzzy feeling out of expanding my knowledge, but if you're going to displace that motivation with only a chance at a measly $10 I guess it's not worth my time.
Here's an example where large potential incentives have failed to override bias-- I think it's an example of what you're looking for.
Good clothing for fat women has very limited availability.
You might think that the unending publicity about people getting fatter combined with a modest amount of observation that people like dressing well would lead to the conclusion that there are a lot of people who'd pay plenty, but it doesn't seem to register.
Or were you looking for explicit incentives which don't require effort to notice?
A different case of incentives failing to work: People haven't gotten better at avoiding market bubbles.
What I want to know is why no one sells half-bras. There's a market: most women are at least somewhat asymmetrical, plenty by enough to warrant different cup sizes. It wouldn't be revolutionary bra technology: it would just have to fasten in the front and the back both and be packaged individually. And it wouldn't take up much extra store space to stock the same range of sizes. I looked once, and there's a patent on it, but no one seems to actually manufacture the things.
Most people have different sized feet and shoes are already separate yet shoes are sold in pairs of matching sizes. I suspect that if you can figure out why that is you will also gain insight into the bra question.
But it seems an obvious market failure that I can't go into a store, pick out a left cup and a right cup and the straps of my choice, and walk out with something that will work better for me than anything I could find in Target without significant extra expense.
OTOH, look at the signaling implications of such a purchase. There's a big difference between knowing you're asymmetrical, and going and buying special clothing because of it. Sure, some people will buy it, but it seems unlikely to achieve mass acceptance.
To the extent that 'good' equals 'fashionable' for clothing I suspect this is a harder problem than it initially seems. What is fashionable is largely defined by what trend setting / fashionable people are currently wearing and the set of trend setting / fashionable women is almost entirely disjoint from the set of fat women. Therefore styles and cuts that are designed to flatter fat women will never become fashionable and fashionable clothing will not scale up well to body shapes it was not designed for.
If you actually wanted to address this issue you'd have to make fat women into trend setters which is a much harder problem than simply scaling up fashionable clothing. The market need you have identified is largely targeted by the dieting and weight loss industries which are very large and profitable.
Status is part of it, but there's a perfectly good statistical explanation too.
There is much higher variance among fat people than among thin people. It's the long tail of the distribution. So plus sizes are much more approximate. It's more likely that the clothes won't fit. This also makes the return on each additional size lower -- there may be a lot of plus-size women generally, but they're spread out enough that there aren't a lot of size 16s specifically.
I don't think that accounts for everything, but it is part of it.
You're already seeing more good plus-size fashion, I think, of necessity. It's coming.
Incentives do not operate by magic: they work by focusing attention and by prolonging deliberation. Consequently, they are more likely to prevent errors that arise from insufficient attention and effort than errors that arise from misperception or faulty intuition.
-- Kahneman & Tversky (1986), Rational choice and the framing of decisions, pdf
If you're making a judgment using a process that would become more accurate (less biased) with more attention, effort, or thought, then incentives could help. And if you have some awareness that you're biased (...
I'm going to bounce a theory off you guys: the reason sports and entertainment have integrated faster than finance is because people care more about sports and entertainment.
I consider entertainment to be a field with a substantial degree of objective measurement because people are good at telling whether they've been entertained.
I would expect things like being a cognitive miser or beliefs one professes for signaling to go away under strong enough incentives, but not mindware gaps.
The difference is that in some problems people just have no idea how to get the right answer (eg talking sanely about GAI), and in other problems, they choose not to for some reason (eg it sounds better to say that you're a better than average driver than a worse than average driver)
The candle problem may be one of the background ideas which contribute to the belief that large incentives don't change biases, even though it's a loose connection-- it shows that incentives reduce creativity.
I think I've found the source of the meme here Dan Pink on the surprising science of motivation.
I'm not sure it quite proves everything it wants to prove-- big rewards do induce creativity among scammers, even if it's mostly variations on a few themes.
A lot of creativity went into creating the financial crisis-- I'm not sure how much of it was deliber...
When you want to get new complex behavior, there's a lot of evidence that it can be shaped with small rewards for changes which more and more closely approximate the new behavior.
This is very different from a great big reward for something which is way out there in possibility space.
This may be a variation on loss aversion-- most people aren't willing to put out a lot of effort unless they're sure of what they'll be rewarded for.
Natural financial (and non-financial) incentives, by which I mean those incentives that are part of the model in which you're observing bias, do probably reduce bias, by letting less-biased actors win more. Artificial incentives from outside the model can alter behavior, but not directly change the rationality level.
Many biases are self-serving enough that they'll alter to match the behavior that outside incentives cause. I'm not sure that counts as a reduction in bias, or just a shift to biases that the incent-or prefers.
The most obvious example of incentives overcoming bias (in the colloquial sense, which I think includes the sense you mean) is blacks and sports in the US. It eventually became clear that excluding excellent athletes didn't make financial sense.
As an incentive, I'll paypal $10 to the commenter whose answer is least biased and most useful.
Do you have some objective method to determine biasedness and usefulness of the answers?
I think Vassar made this argument (incentives do not fix bias) based on the assumptions that
One of Tyler Cowen's recent books, probably "Discover your Inner Economist" discusses this in depth.
Some other experiments, possibly from K & T, show the same irrationalities among very poor third world residents when dealing with amounts of money that are very substantial to them.
I'm trying to better understand the relationship between incentivization and rationality, and it occurred to me that it is a "folk fact" around here that large financial incentives don't make cognitive biases go away.
However, I can't seem to find any papers that actually say this. It's not easy to google for (I have tried) so I wonder if the Less Wrong collective memory knows how to find the papers?
Is there a pattern to which biases go away with incentivization? Do we have at least 5 examples of biases that go away with incentivization and 5 examples that don't go away with incentivization?
As an incentive, I'll paypal $10 to the commenter whose answer is least biased and most useful.