Tuesday, July 08, 2014

What We Could Learn From Dan Kahan. And Perhaps Shouldn't

I have commented in earlier posts on Dan Kahan's research about why we believe what we do. The basic conclusion is that, in the case of beliefs linked to group identity, individuals choose what to believe not on the basis of the evidence but on the basis of what beliefs fit their group identity. For most of us, as Dan points out, that is rational behavior. My beliefs about evolution or global warming have  little effect on the world at large but could have a large effect on how well I get along with the people who matter to me.

There is no reason to assume that the pattern I have just described only applies to other people. One possible conclusion is that I ought to look for beliefs that it is very much in my interest to hold and think about how good my reasons are for holding them.

One of the first candidates that occurs to me is my belief that minimum wage laws have bad effects, that they make it harder for unskilled workers to get jobs. That is a belief widely held among economists, especially economists generally in favor of free markets. Academic economists are part of an academic world where most people are politically liberal, hence inclined to support minimum wage laws. That makes the fact that economists are inclined to oppose them an important marker for group identity. And I have evidence of the cost to an economist of rejecting it from an interview with David Card, coauthor of a widely cited study that failed to find the expected effect. By his account, getting what was, from the standpoint of his professional peers, the wrong answer cost him a lot of friends.

That is a reason to think that my view on the subject might be mistaken but I cannot think of any good arguments against it, so I will rapidly pass on from my possible false beliefs to other people's. Dan Kahan's, for example.

Dan is a professor at Yale law school. My guess is that if he pointed out that there is no good reason to expect the distribution of intellectual abilities to be the same for men as for women, hence no good reason to interpret unequal outcomes as evidence of discrimination, he would suffer significant social costs. If he made the same point in the context of racial differences, the costs would be more than merely significant. Perhaps he should think about those questions, examine the arguments and evidence, and revise his beliefs accordingly.

Or perhaps not. His explanation for what he observed, after all, was that it was the result of rational behavior. Believing things merely because the evidence supports them may sound like an admirable policy, but it could prove to be a very expensive one.


24 comments:

Mike Fagan said...

Yet convictions* that are widely held today may not be widely held at some point in the future. Things change. It may be that holding and publicly maintaining a belief in a proposition that is contrary to prevailing convictions may eventually pay dividends of one sort or another. A loose analogy might be to the idea of stock market "bubbles", and those who either "predict" the bursting of the bubble or sell the stock at peak prices before the "burst".

But obviously, assessing which prevailing convictions might change and when might be easier said than done, and nor is it certain what, if any, the possible "dividends" to publicly maintaining a contrary belief may be, or on what timescale they may occur. So some people might decide to keep certain beliefs (or even just conjectures) to themselves, whilst others go all out in publicly maintaining them.

*I hesitate to call them "beliefs", given that holding them may demand the cognitive evasion of falsifying evidence.

Daublin said...

The examples you give do matter, though. Just not to the speaker. If we *all* just fiddle away with our friends, then we will win individually but suffer as a group.

There are other examples where the questionable belief doesn't cause harm to either the speaker or to the human race. For example, does it matter if we think people have souls? I am hard pressed to think of where it might matter.

Anonymous said...

A belief that people have souls can matter. In particular, it can motivate the believers into attempting to "help" or "save" those souls at the expense of the victims' material well-being, or even of the victims' lives.

gurugeorge said...

It's incredibly hard to pull away from the consensus of one's peers. I was a socialist, verging on communist (contemplated joining the CPGB at one time), and all my friends were strongly socialist.

Unfortunately, or fortunately, depending on how one looks at it, I had had ingrained in me the idea that I should try and understand the enemy in order to overcome them, and as I started reading the enemy, I started to become more and more convinced first, that the enemy wasn't all bad, next that the enemy occasionally talked some sense, next that the enemy was actually talking considerable sense, and finally, that it wasn't the enemy, and that my former beliefs were the enemy. Further maturity has led me to think that enemy-talk isn't actually conducive to the advancement of anything.

But at any rate, during this learning process, it was hard to stand on my own. At first I had a sense of cognitive dissonance as I would agree with my friends at table talk, but secretly be thinking "no, that's not right". Then I started to timidly voice objections, and suffer pitying looks and "tsk tsk"s. Then I started to argue, and get more oppositional, to evident bafflement. Now, I'm the crazy "right" winger amongst my friends. Everything that made friendship is still there, they just think I'm mad, and that at some point I'll get over it.

So yes, the cost is great, and it's often tempting not to pay it.

There's an analogous syndrome wrt philosopher Daniel C. Dennett's (and his research partner's, forget her name) investigations into closet-non-believing clergy - fascinating to read about.

JES said...

"there is no good reason to expect the distribution of intellectual abilities to be the same for men as for women."

Actually, there is a good reason ... science has thoroughly debunked the outdated ideas that there is any such difference in intelligence:

http://www.apa.org/monitor/jan06/gender.aspx

And with greater gender equity, the gap between male and female performance in math and science shrinks:
http://scienceblogs.com/startswithabang/2012/01/06/surprise-surprise-gender-equal/

But by all means, don't let real science get in the way of a good essentialist narrative.

macsnafu said...

On the one hand, this is very enlightening, and explains much about political arguments. On the other hand, this seems quite trivial, and probably ought to be self-evident to most people (even if it really isn't). Nonetheless, I have to believe that the reality and truth of the situation has to matter at some point.

The minimum wage, for example, is enforced by law, and I don't care how many people support it, if it does economic harm, then it does economic harm.

Or the global warming issue. It not only matters if one's view of GW is right or wrong, but, given that there are multiple positions on it, not merely for or agains, it matters how closely one is right or wrong.

How do we determine "right" and "wrong" when it comes to complex issues? Economics would seem to provide invaluable feedback in this regard. Is recycling good? Well, not if it costs more than not recycling. And that can change over time, sometimes very quickly. Even if I don't collect aluminum cans, I like going by the local metal collection business and seeing how much they're paying for aluminum this week.

But with government-mandated recycling, it becomes harder to determine when recycling is actually worthwhile, although one could always say that if it has to be government-mandated, then it's not worthwhile yet.

Like recycling and the minimum wage, the great danger with the global warming issue is that governments will step in and force people to do things that make little or no economic sense, distort the marketplace and make it ever harder to know when someone is doing something "right" or "wrong".

Tibor said...

JES:
First of all, sorry for the length...this is divided into two separate comments, it was too long for one.

Your first comment is just a summary which does not provide any arguments or statistics. It mentions an article that supposedly does, but on its own it is not helpful at all.

Also, the problem with the field of psychology is that it is inherently very biased. A vast majority of psychologists are of social democratic "liberal" persuasion, a handful are libertarian-ish and even fewer are republican "conservative". That is of course not an excuse to dismiss good arguments, but they have to be present first and it is a reason to be careful. If someone has an axe to grind, he is more likely to try too hard to get the results he wants to get (which, of course, does not apply only to psychologists or a particular political mindset).

Now, one psychologist who actually studies these biases and also questions like "Why do good people diverge in political opinions?" is Jonathan Haidt (by his words a "liberal turned centrist"). He essentially wants to find out what causes the different political views of people. You might like what he does.

The second link:

Now, this is better, as there are some arguments, based on some statistics.

Some comments:

The author argues that the variance between men and women is the same and supports his claim with some histograms. Those histograms are of students who take part in the mathematical Olympiad at school and their outcomes at the Olympiad (presumably, it is not stated there clearly and the x-axis is not labelled). That is hardly representative for the whole population. First of all, most of those who participate there at all are already good (compared to their peers) at maths. At least in the Czech republic (where I come from), participation is not mandatory at school. That ignores all the rest and the variance is just a variance among the very good students, not variance of skills overall. Secondly, the success at the mathematical Olympiad and similar tests are quite heavily dependent on how much you study. That is not necessarily the same as how talented you are. The common pattern is that girls at school study much harder than the boys overall, since boys are generally more restless. So this might be a test of skills, but not of talents.

I should add that I really don't like the high and mighty and "righteous" tone of the article, but that is mainly a question of taste, I suppose.

Tibor said...

JES:

Here it continues...



Still, I think that while there may or may not be difference in the variability of talent, there is much to say about differences in goals and desires.

I studied maths in Prague (now I'm a PhD student in Göttingen) and it was interesting to observe the differences in sex among different sub-fields. Essentially, the closer you got to applied fields (such as econometrics or financial mathematics), the higher the percentage of female students. I believe that in our year, the financial maths had either the equal number of female and male students (and there are, fortunately, no sex quotas in the Czech republic), perhaps there were even more female students. In more theoretical fields (algebra for example), there were almost no female students at all. And it was not even clear that the less talented people went for the applied fields. Rather those were more, let's say, practically oriented people. If you go for theoretical maths, you face a lot more uncertainty in your future career than as an econometrist (I should say that you can usually still get the same jobs anyway, with some additional work of learning a few extra things, but it is definitely not the easier way). From the evolutionary perspective, there are good reasons to expect women to be more conservative, that is to say, to go with the less risky options which do not have the nice property of making you incredibly successful if you are very lucky, but at the same time have the very nice property that your chances of ending up like a miserable sucker are far lower also. Similarly in business - fewer women are willing to stay at work 11 hours a day to raise the corporate ladder (not many men are willing to do that either, but of those people who do, most are men). Or, to take an example from history, never has a group of women sailed to plunder or explore foreign lands so that a handful of them would return with the spoils and glory. They would have not much to gain by that and a lot to lose. The man who stayed at home however, would be at a serious disadvantage on the dating market, so to speak. Women usually don't like losers, while men usually do not (mainly) care about status in their partner. So a woman who stayed was fine and avoided the high chance of being killed in a foreign land. A man who did perhaps survive, but his chance with women were rather bad.

So you can have women as talented as men, even with the exact same distribution (or with just a slightly less variance as the article suggests, so that it cannot explain the differences in outcomes by itself) and still observe these differences, even in a society where sex is simply not an issue outside of personal relationships (and some obvious exceptions such hiring a ballet dancer or a bouncer at a club). I believe that the differences in goals and desires are what causes these differences more than anything else.

Also, what definitely has to play at least some role in the differences in outcomes is that women give birth. That means fewer years at work overall and it also means a higher uncertainty for an employer who considers hiring someone to a position (basically, you don't want your crucial employees to be gone, even for half a year, when you need them...and most women take a longer maternity leave than that). You may not like that, but it is just a fact.

There are of course exceptions, you can point out to very successful female businessmen (businesswomen?) or scientists (by the way, my PhD advisor is a woman and I believe that she is quite brilliant...a lot more talented than me). But if women are indeed overall less ambitious (in the career related things) than men, then you cannot expect the same results. It is sort of like with sports. I don't believe Americans are worse suited to play football (soccer) than Germans, but Germans like it more and so more Germans play the game and if you pick great football players randomly, more of them will be German than American even though overall Germans and Americans can be equally talented.

David Friedman said...

Let me generalize Tibor's argument:

Reproductively speaking, being male is a high risk profession. A successful male can produce a very large number of offspring, an unsuccessful one none.

Females possess the scarce reproductive resource: Womb space. Hence even a not very successful female can expect to reproduce, and a very successful female is limited to the number of children she can bear.

To maximize the return in a high variance outcome gamble, it is worth taking risks in order to have a chance of ending up with a very high payoff result. Hence we would expect evolution to produce a wider spread in the distribution of abilities for males than females. I believe that prediction is consistent with observation on things such as IQ distributions.

Shaddox said...

"To maximize the return in a high variance outcome gamble, it is worth taking risks in order to have a chance of ending up with a very high payoff result."

Forgive my ignorance of biology, but why is that true? Looking at the problem from the perspective of statistics rather than biology, it seems to me that you would want to maximize expected value rather than variance.

By definition, I would expect a high-variance and a low-variance model with the same expected value to perform equally. And I would expect a high-variance model to perform worse than a low-variance model with slightly higher expected value.

James A. Donald said...

Ghenghis Kahn had several thousand children. No Queen, no matter how powerful, can have more than a dozen or so children. So natural selection favors men trying to conquer the world, and women trying to clean their kitchen

So natural selection favors women's brains in which nothing goes seriously wrong, and men's brains with the potential for genius or madness.

So women will be closer to the mean, and that mean slightly dimmer on average, though more practical on average. Women are a lot better at finding your keys than you are.

So any profession in which the minimum requirement is substantially more smarts than the mean, will necessarily have substantially fewer women than men competent to do it.

David Friedman said...

Shaddox:

You want to maximize expected return. But if high end outcomes produce very large gains and low end outcomes can't produce less than zero, a high variance strategy gives you a high expected return.

Dan Kahan said...

how about that Bucky F*ing Dent's "homerun" was fair ball? I've watched the tape 500 times--it was foul!

I am sure I have views that reflect the dynamics I'm studying. I can't tell which ones, of course.

I doubt that I'm special in either of these regards. Or in being eager to figure out how to counteract the incentives that *make* it rational to attend to information in this manner

Tibor said...

Shaddox:

Also, this is basically answered by David's previous post - "improving is sometimes hard". If you could just improve overall, you'd do that. Or more precisely, the evolution would do that. People (or any other organisms) who could just be better overall, would eventually dominate in the population. This is in population models called a selective sweep. One actual example of this is the gene for digesting milk (as an adult). That simply makes you "better" (as in "more fit to the environment you live in") and that is why it spread quickly through (almost) all humans. You still have mutations, so the gene is really not present 100% of the time, but a vast majority of people have that. (outside of remote and isolated populations where the original beneficial mutation never happened, or it did but did not spread and died out since they were not herding cattle and therefore it did not increase fitness in such an environment)

So there is a good reason to expect the means of talents to be "optimized", in other words, the only way for you to get more very fit individuals is to increase the variance (the means are already maxed out by evolution). Now it seems today, that (based on inference on the DNA of today's population) in all of the history, about 80% of women reproduced and their genealogies survived until today, compared only to only 40% men for that number. So being a men used to be quite a high-rollers game, an all-or-nothing situation in which it is best to take a lot of risks. You will most likely lose, but if you are "conservative" and take few risks, you lose (in reproductive terms) anyway. If you are a woman, your chances are pretty good though and taking high risks is a bad idea.

One could argue that if this were true, women would be on average "worse" also, since the competition is not as high among them...Now, while it is true that a woman with really good genes overall will have it harder to spread her great genes compared to a man with really good genes (at least historically, maybe not to much in a monogamous society where it is hard for both) since she can only have much fewer children than the man can, daughters inherit their fathers' genes also. So there is probably not a good reason to expect the average talents to be different and there does not seem to be much evidence that they are different either.

In a monogamous society, these differences could eventually go away though. If being a "loser" is no longer a problem and you still have a good chance to reproduce anyway, the variance in men might shrink. But evolution is a very long process and monogamous society is not older than about 2000 years (and much younger in most of the world). Before it could take any measurable effects (another couple of thousands of years), we will probably have ways to improve our genes overall by genetic engineering (evolution works in small steps - if something is really beneficial after 3 steps, but the first two make you considerably less fit, then it does not happen. With genetic engineering, if you understand how the genes work well enough, you can bypass this limitation...), or alternatively, there might not be any humans left anyway...I think the first option is still more likely.

David Friedman said...

Dan:

I think there is a way of spotting candidates for views held for that reason, especially if the views involve believing in bad things happening, such as CAGW. Imagine that you come across evidence or arguments suggesting that AGW isn't really a problem. How do you react?

On the face of it, you should be happy, hope the evidence is right, look forward to telling your friends about it.

My guess is that most people in our social environment would instead hope the evidence was wrong and look for reasons to reject it—because they could imagine the social consequences of announcing to their friends that they no longer believed warming was a problem.

So use introspection to find issues where you would have the latter reaction to what ought to be good news.

David Friedman said...

Tibor writes:

"So there is probably not a good reason to expect the average talents to be different and there does not seem to be much evidence that they are different either."

I'm not sure what "average talents are different" means in this context. Once you have different variances, whether the average is or isn't the same depends on the particular metric you use, and I don't think there is an obvious one.

I'm not sure it's true, but I believe I did read that IQ tests were at some point designed to give the same average result for men and women by suitable weighting of questions on which one gender or the other did better. Clearly it could be done, and makes the claim that average IQ's are the same (or are different, unless the difference is very large) meaningless.

Tibor said...

David:

I'm not sure I understand. What do you mean by the dependence on a metric? That which measures the "skill levels" in individuals and assigns a number to them?

By the way, I'm not sure IQ is a very good measure of intelligence (whatever intelligence is). Mainly because that if you do a lot of IQ tests, you improve your IQ score, but being familiar with certain types of questions (adding a right number to a sequence for example) does not count as increasing your intelligence (which I understand as a potential, or a talent, not a skill). Also, sometimes the tests implicitly require certain knowledge - such as a sequence of prime numbers where you are supposed to fill in a missing prime. If you have never heard of prime numbers, you are unlikely to figure it out. If you have, it is trivial.

I'm afraid that in the assessment of intelligence, one has to stick to metrics that give less precise results but give them more correctly. Such as judging the work of someone or generally an impression one has from talking to him.

Joey said...

Tibor:

I'm not sure I understand. What do you mean by the dependence on a metric? That which measures the "skill levels" in individuals and assigns a number to them?

I'm guessing he means that if there were some objective average intelligence that was identical between men and women, every real-life IQ test will produce different averages if the male and female variances are not identical.

If a test is devised such that you need to be in the top thousandth of intelligence to get any questions right, if male variance is higher, then the male mean will be higher.

If you design an IQ test that is just hard enough that you need to be in the top 99.9% to get any questions right, then the female mean will be higher.

It's impossible to design a fair IQ test. This is why they weight the questions after the fact to force the male and female means together.

Tibor said...

Joseph: That's right. But if you make the test so that all the difficulty levels are uniformly distributed, you don't have to worry about the differences in variance.

I concede that this is practically impossible to do, so yours and David's point is good (and it is interesting that despite me studying probability theory, it did not strike me as an obvious problem...I guess this is what you get if you do the theory and not concern yourself with real world implementations :) )

Anyway, this then is another reason for me to frown upon IQ tests (and especially people obsessed with them) :)

starrychloe said...

Ideologies are genetic. Even if you have been brainwashed in an opposing idealogy, you will gravitate to the one you are genetically predisposed to once you become aware of it.
http://reason.com/archives/2012/04/10/born-this-way/print

Tibor said...

Unknown: Check work of Jonathan Haidt. He is a psychologist who studies the reasons people hold different political opinions.

I am a bit skeptical to the claim that ideologies are determined by genetics...the complex societies of today have not existed until some 3 thousands years back (and those were still far simpler compared to ours). Evolution does not work that fast.

But I have not read the article yet (I am in a bit of a hurry to catch a train and probably should not even be writing this now :) )

Anonymous said...

It's also important to note whether a minimum wage is binding, and if so, to what extent.

Benjamin Cole said...

I always enjoy DF's posts. If only my liberal and conservative friends could be as open-minded.

Anthony said...

JES - your first link doesn't even address variance. Your second link shows a consistently higher variance among men across ethnicity, educational styles, and testing methods.

David - regarding the minimum wage question, it's probably fairly complicated, and economic and political conditions in the U.S. and other developed countries probably limit most of the damage that a politically-feasible increase in the minimum wage might cause. Hypotheses for investigation:

1) The harm isn't in immediate unemployment - it comes from slower rates of pay raises, or lower corporate profits causing lower job growth beyond the time horizon that almost any study can see direct effects.

2) There's significant evasion of minimum wage laws - raising it to the point of unprofitability for a company leads that company to be replaced by one which uses contract labor or illegal immigrants or family members working at effectively less than the legal minimum. Some people lose jobs, others gain jobs at the same, now illegal, rate of pay.

3) Under current labor market competition conditions, employers have such an advantage that they can bid low-skill labor to accept pay significantly lower than their productivity; increasing the minimum wage laws make only a trivial number of jobs cost more than the new minimum, and so the main effect is to shift the rewards production a little from capital to labor.