Today's post, Rational vs. Scientific Ev-Psych was originally published on 04 January 2008. A summary (taken from the LW wiki):

 

In Evolutionary Biology or Psychology, a nice-sounding but untested theory is referred to as a "just-so story", after the stories written by Rudyard Kipling. But, if there is a way to test the theory, people tend to consider it more likely to be correct. This is not a rational tendency.


Discuss the post here (rather than in the comments to the original post).

This post is part of the Rerunning the Sequences series, where we'll be going through Eliezer Yudkowsky's old posts in order so that people who are interested can (re-)read and discuss them. The previous post was Stop Voting For Nincompoops, and you can use the sequence_reruns tag or rss feed to follow the rest of the series.

Sequence reruns are a community-driven effort. You can participate by re-reading the sequence post, discussing it here, posting the next day's sequence reruns post, or summarizing forthcoming articles on the wiki. Go here for more details, or to have meta discussions about the Rerunning the Sequences series.

New Comment
22 comments, sorted by Click to highlight new comments since: Today at 11:48 AM

One of the major challenges to evo-psych hypotheses about gender is never really tackled in the original comment thread: that women and men are conditioned differently in almost all current and historical societies, so it's almost impossible to differentiate effects of training from inborn psychological phenomena.

Basically: if you're the one who's always been expected to put the groceries away, do the cooking, and set the table, you've probably developed a pretty good strategy for getting the damn ketchup out of the fridge.

Of course, we could always argue that women are almost always socialized for these roles because they have an evolution-granted knack for them, but then we'll be tasked with finding a large enough (and representative enough) population that hasn't had any of that conditioning.

One of the major challenges to evo-psych hypotheses about gender is never really tackled in the original comment thread: that women and men are conditioned differently in almost all current and historical societies,

This fact (or rather the fact you seem to mean, i.e., that the differences have a similar character across almost all societies) is itself evidence for the evo-psych hypotheses about gender.

Even if you assume that societies are more likely to structure their gender constructs around innate psychological traits than physical traits (or that the former would result from the latter), you've got a major confound when assessing the strength of the effect.

It's not difficult to show that biological sex predicts some features of perceptual/cognitive ability. It is difficult to show that people of a gender that is generally trained to work with refrigerators have, as an innate feature of their psychology, an effective strategy for searching fridges.

Even if you assume that societies are more likely to structure their gender constructs around innate psychological traits than physical traits

Eugine's argument was that the probability of societies structuring their gender constructs around innate psychological traits has to be updated upwards on seeing culturally universal gender stereotypes that have something to do with psychology. It doesn't need assumption of high prior on the hypothesis.

you've got a major confound when assessing the strength of the effect.

That's true, but irrelevant with respect to the parent comment.

It is difficult to show that people of a gender that is generally trained to work with refrigerators have, as an innate feature of their psychology, an effective strategy for searching fridges.

Do the test with cooks of both genders; their experience of using fridges is unlikely to differ significantly in length. Do another test with women raised in feminist families and compare to general population.

I didn't dispute Eugine's argument - I just thought it worthwhile to point out that the evidence itself is obviously confounded. If we consider the confound itself - the gender-based training - evidence of the hypothesis, we're stuck in a tricky chicken-and-egg situation. It wasn't a refutation of Eugine's comment, but I hardly think it's irrelevant.

Do the test with cooks of both genders; their experience of using fridges is unlikely to differ significantly in length.

Unless female cooks are more likely to become professionals as the result of early and consistent pressure (as opposed to other motivations) and more likely to do the grocery shopping/cooking at home, etc. You can try to control for gender conditioning, but it's pervasive enough to be a significant challenge.

Do another test with women raised in feminist families and compare to general population.

I'm not sure 'feminist household' is equatable to 'egalitarian household,' in practice - but even if it were, self-identifying as feminist is not the same thing as somehow overcoming all early gender conditioning.

If we consider the confound itself - the gender-based training - evidence of the hypothesis, we're stuck in a tricky chicken-and-egg situation.

I am not sure if I understand what exactly is tricky in this situation. I assume that "the hypothesis" is something like "the innate psychological differences between the sexes are the main basis for the gender constructs". There is some trickiness associated with the exact meaning of the hypothesis, but no trickiness with the evidence. Remember that the evidence is not the sole existence of gender-based training, but the fact the training is culturally universal. If the division between the roles of men and women were purely arbitrary or based on physical differences, one would expect more variability between cultures; some cultures would have men in kitchens.

Unless female cooks are more likely to become professionals as the result of early and consistent pressure (as opposed to other motivations) and more likely to do the grocery shopping/cooking at home, etc.

If women don't stare at open fridges because they are using them every day and have thus have learned how to search there efficiently (that was your suggested hypothesis), then male cooks will also not stare at open fridges because they use them every day. Perhaps female cooks are even better, but if the test is done and it is found that male cooks who spend on average 5 hours a day cooking stare at the fridge significantly more than female non-cooks who spend on average 2 hours a day cooking, there goes your theory.

I'm not sure 'feminist household' is equatable to 'egalitarian household,'

Be it egalitarian.

in practice - but even if it were, self-identifying as feminist is not the same thing as somehow overcoming all early gender conditioning.

Therefore I haven't suggested doing the fridge test with self-identified feminists, but with women raised in feminist (or egalitarian, if you wish) households. Or, more directly, simply select women who never cook or prepare food. You will certainly find some.

In general, I don't say the evidence in any of the discussed examples is strong. But I object to your implicit insistence that it is virtually impossible to test hypotheses about inherited psychological differences between the sexes.

Well, it's evidence that there is some difference between men and women that exists throughout all societies. I think the main epistemic problem is that the explanatory power of the null hypothesis (men and women are natively psychologically almost identical, but people know that women get pregnant and men don't and social structures take this into account) is pretty high and, even worse, often gives reasoning that's isomorphic to ev psychic explanations.

This isn't exclusively a problem for gender, either - oftentimes ev psychic explanations in other domains tend to give explanations that are isomorphic to how one would explain the behavior of a minimally modular rational actor.

Of course if Eliezer is correct - and as best I can tell his reasoning is sound, although the conclusions, even when transported to much less mindkilly domains, seem to be absurd - then this isn't much of a problem at all.

I think the main epistemic problem is that the explanatory power of the null hypothesis (men and women are natively psychologically almost identical, but people know that women get pregnant and men don't and social structures take this into account) is pretty high and, even worse, often gives reasoning that's isomorphic to ev psychic explanations.

There's still a noticeable amount that it doesn't explain. To pick just one easy example: men and women employ different strategies for navigation. There's no particular reason in modern society for that to be true. If those traits were entirely learned, one would not expect men to have, purely by coincidence, learned only the skills that corresponded in the ancestral environment to increased navigation skills while hunting and at war, and women to have, purely by coincidence, learned only the skills that corresponded to navigation skills necessary for gathering.

More generally, a psychological trait which

  1. is useful for purposes present in the ancestral environment,
  2. is not useful at all in the present society, and
  3. most people are unaware of it

is much better explained as inherited than by cultural conditioning.

From someone who believes cultural conditioning is more powerful than you seem to, here are some (hopefully) constructive criticisms, in the interest of sharpening the disagreement.

most people are unaware of it

Cultural conditioning asserts that the conditioned traits feel "natural." At the very least, unawareness of the traits should not be evidence against cultural conditioning.

is not useful at all in the present society

"Not useful at all" seems like a great candidate for typical mind fallacy. Cf. status politics. I suggest a friendly amendment like "not presently used for the same purpose as the ancestral environment".

More generally, a psychological trait which is useful for purposes present in the ancestral environment and not presently used for the same purpose as the ancestral environment is much better explained as inherited than by cultural conditioning.

That seems closer to true. Does it omit any important cases?

who believes cultural conditioning is more powerful than you seem to

How much powerful I seem to believe cultural conditioning is? I suspect that you underestimate my trust in the power of culture.

Cultural conditioning asserts that the conditioned traits feel "natural." At the very least, unawareness of the traits should not be evidence against cultural conditioning.

I didn't mean unaware of the fact that the trait is result of cultural conditioning, but that the trait exists at all. If, for example, nobody knows that women write on average 20% smaller letters than men (I made this up), then how the society manages to train women into smaller hadwriting?

But we are discussing socially contested positions (like possibly sex-based differences in ability to do math). What does it mean for the society to be unaware of the trait, when the very dispute is whether the trait is inherent or cultural? By contrast, the existence of cognitive bias is not culturally contested, which is some evidence that it is inherent rather than culturally conditioned.

nobody knows that women write on average 20% smaller letters than men (I made this up), then how the society manages to train women into smaller handwriting?

I'm sorry, I can't parse this.

This is outside of my area of expertise, so I won't offer judgement on the reliability of the metastudy referenced in the link, but not only is the biological explanation for gender differences in math performance contested, the existence of the phenomenon itself is apparently contested as well

The society in general can be unaware of a trait while few people who are aware of it (perhaps those who have just discovered it through a psychological experiment) can discuss its origin. I don't see any contradiction in this. Of course it doesn't apply to sex-based differences in math proficiency which are widely discussed. It can, on the other hand, in principle apply to the refrigerator staring problem discussed in the OP (although, as it has been pointed out, this exact difference can be caused by differences in gender roles with respect to preparing food).

To expand the unparseable section: Imagine you make a study to compare handwriting of men and women in which you, possibly among other things, measure the size of the lettres. You realise that female handwriting is smaller on average by 20% than male handwriting. You repeat the study to exclude confounding factors (compose your experimental groups of men and women of the same education, social and ethnic background, age, handwriting practice, physical dimensions, whatever else) and see that the effect persists. The main remaining explanations are

  1. Men have inherent psychological inclination to write bigger lettres than women.
  2. Men are socially conditioned to write bigger lettres.

Now imagine that the society is unaware of the fact. That means: if you ask random people whether men or women write bigger lettres, most people would answer "I don't know" and the rest would split equally between the other possibilities; nobody would refer to small compact handwriting as feminine and to big, crude letters as manly. This leaves 2 with a problem: social conditioning is supposed to work by making gender stereotypes apparent and people then accomodate to the roles they are expected to play and traits they are expected to have. But when the traits are not apparent, this couldn't work. Which pretty leaves us with 1 as the only explanation.

(Note that I have chosen a fictitious example because I am not an experimental psychologist and so I lack knowledge of sex-based differences or other psychological traits unknown to general public.)

I think the main epistemic problem is that the explanatory power of the null hypothesis (men and women are natively psychologically almost identical, but people know that women get pregnant and men don't and social structures take this into account) is pretty high and, even worse, often gives reasoning that's isomorphic to ev psychic explanations.

So your null hypothesis is that genetic evolution failed to produce any traits in humans that accommodate this fact, so all behavioral differences are the result of memetic evolution and explicit reasoning. This hypothesis strikes me as extremely unlikely.

Not that I disagree with you, but I've just this second noticed how dangerous "almost impossible" is as an expression.

There is possibly another gender based refrigerator difference besides men waiting for items to appear and women sorting through them. I think men return to refrigerators to look for food more often than women do, after having been disappointed by the selection an hour or so ago, just in case something appeared in the interim.

This is testable! Have volunteers wait in a room "for the experiment to begin." Tell them they may eat anything in a refrigerator. Have nothing but crappy food in the refrigerator. See who, after opening the refrigerator and not taking anything, opens it again without taking anything.

What I found interesting about this post was the increased risk of the mind projection fallacy- because he knows about evo psych, any hypothesis he can construct that conforms to the evo psych narrative gains credibility, even without data to support it. That is, he makes a conclusion about male and female strategies from a sample size of n=2. Is it more likely that a sample size of n=200 will repeat that result than have the opposite result? Yes, but not by all that much- and so I would suggest more reluctance at identifying one of his traits as a male trait.

Specifics aside, love this post. This was an area of strongly felt confusion for me.

To a first approximation, I agree with the original essay. To nitpick, though:

  1. If someone has thought through his model in enough detail that he has a reasonable-sounding plan to empirically test it, that's bayesian evidence that it's a strong one.

  2. If someone's model has no in-principle testable consequences, that's bayesian evidence that there's little point in wasting brainpower considering it; scoffing at it and forgetting it is a safe default action. Of course, since the ketchup hypothesis does have such consequences, I can think of uses for it, if it is true.

If someone has thought through his model in enough detail that he has a reasonable-sounding plan to empirically test it, that's bayesian evidence that it's a strong one.

Upvoted this ^^^

If someone's model has no in-principle testable consequences, that's bayesian evidence that there's little point in wasting brainpower considering it; scoffing at it and forgetting it is a safe default action. Of course, since the ketchup hypothesis does have such consequences, I can think of uses for it, if it is true.

What about the case where the model is not testable, but possibly actionable in your lifetime, e.g. asteroid strike in 1K years prevented by an slight orbit deviation produced by a nuke. I'm guessing this is covered by the in-principle clause.