Do you think "rationalists", or rationality leads people to, stereotype demographic groups less or more than the general public?

post by orthogenesis · 2021-06-06T16:50:25.916Z · LW · GW · 9 comments

This is a question post.

Contents

  Answers
    2 Viliam
None
9 comments

For many conventional things that are considered standard "stereotypes" about groups such as physical appearance-related assumptions about sex or gender, age, race, nationality, or style of dress, accent etc., that the general public employs?

On the one hand, those in the rationality community emphasize the point to use whatever information possible, and that means sometimes using statistical info that would be called stereotyping (the famous example of statistical discrimination being car insurance for men and women, and other practical things such as avoiding tall men in dark alleys but not short women). 

However, rationalists often also emphasize updating priors so it could be that even if initial stereotypes are held, a big deal is made of when to update more readily -- for example, when many people would keep on stereotyping a member of a nationality for e.g. not fitting the stereotype (e.g. are you really an X if you don't look like the X's I know, or have an accent I associate with X, or eat what I think of as X food?), it may be rational to update more readily (I just met a X person, I should update my preconceived notion of what X's are like instead of insisting they're not a real X for not following my stereotype).

Also, self-described rationalists are famously big on individualism (e.g. Scott Alexander's discussion of respecting individual preferences such as introverts vs. extraverts, ask vs. guess culture which is more about individual preferences than traditional demographic categories or cultures). It may be the case that some stereotypes are true on average (80% of X like some food vs. only 40%'s of Y's) but having an individualist, analytic mind (typically seen as more a WEIRD thing or STEM thing or "rationalist" thing, perhaps, though not always) vs. a collectivist, hollistic one might laser focus on the individual's agency and preferences (e.g. I don't care if Bob's a member of group X that stereotypically likes the food, Bob has already told me he doesn't like it, I'll take his word for it).  

Answers

answer by Viliam · 2021-06-06T21:42:06.332Z · LW(p) · GW(p)

You use all data you have. That includes your previous experience with people who have a trait X, your previous experience with people who told you what opinion to have on people who have a trait X, your current interaction with the specific person, etc.

Is it important for you to make more correct conclusions? Get more data.

comment by orthogenesis · 2021-06-06T22:20:49.752Z · LW(p) · GW(p)

I think it goes without saying that more data is good. But the quality or strength of the data is important too. I think some debates over stereotypes rest on if they count as good quality data, or data that should override other data (firsthand experience) on how to update your prior. For instance, if you get data from mass media that "all women like chocolate more than men" but get data from most of the men and women you know that both like chocolate equally, which trumps which in if you are more likely to consider chocolate as gift to male or female friends? 

You could say the societal stereotype is better data -- after all stereotypes have been built up over generations, are "common knowledge". You could say your personal, thoughtful experience is better (I trust my own people around me, not secondhand, thirdhand, or mass media cultural tropes -- but what if I'm in an unrepresentative bubble, what if my friends, knowing the stereotypes are ashamed to fulfil them and say the opposite, in which case I should downweigh their claims and actually follow the stereotype more).

Also in adversarial settings you want to know if stereotypes are accurate data or are created with an agenda (e.g. in wartime many stereotypes about the enemy's traits are not based on accurate understanding of the enemy; okay in a less obviously conflict-driven setting you might get this still -- like stereotypes exaggerated to sell a product "get this for dad" even though your dad doesn't fit the stereotypes, or "this city has friendly people" obviously sponsored by the tourism industry). They could be accurate however and in your best interest (e.g. the stereotype of citizens of this city being mean and unfriendly might be unflattering but your friend might generally care and tell you the stereotype (against the fear of generalizing) because if you're stranded there, it's good to know how much help you can expect from friendly strangers in borrowing a phone).

 

Replies from: Viliam
comment by Viliam · 2021-06-07T19:38:53.322Z · LW(p) · GW(p)

Sometimes there are clever things you could try, for example find out whether female chimpanzees like chocolate more than male chimpanzees... but of course there are situations where the rational answer is simply "I don't know".

That doesn't necessarily mean no data, but could mean data that you strongly suspect are filtered or fake, without being able to sort out this mess. In other words, all evidence you have is very weak evidence: personal evidence may be weak because it is likely to be a result of your bubble (you are more likely to associate with people who like chocolate as much as you do), media evidence may be weak because media do not have sufficient incentives to say true things.

EDIT: Of course, saying "I don't know" can make both sides angry that you don't see how the stereotype of obviously true/false. Sometimes it is smarter to not say what you actually believe, even if the actual belief is "I don't know".

9 comments

Comments sorted by top scores.

comment by ChristianKl · 2021-06-06T17:17:42.986Z · LW(p) · GW(p)

I haven't downvoted this post, but I think the main problem with it is that it doesn't define what it means with stereotyping.

Stereotyping is a highly politically loaded term and one of those model verbs that get used very differently in the first and third person.

Replies from: orthogenesis
comment by orthogenesis · 2021-06-06T17:36:42.285Z · LW(p) · GW(p)

I don't know if wikipedia's entry helps here but hopefully I'll try to formalize or at least give some central examples: 

In social psychology, a stereotype is a generalized belief about a particular category of people.[2] It is an expectation that people might have about every person of a particular group. The type of expectation can vary; it can be, for example, an expectation about the group's personality, preferences, appearance or ability. Stereotypes are sometimes overgeneralized, inaccurate, and resistant to new information, but can sometimes be accurate.[3]

Lee Jussim has published on stereotyping academically and argued (in a way contrarian to the popular consensus) for its accuracy in many settings. For instance "The Empirical Assessment of Stereotype (In)Accuracy, summarizes what is now an impressive body of literature assessing the (in)accuracy of racial, gender, age, national, ethnic, political, and other stereotypes".

Even though it's fuzzy, let's just go with some of the standard central examples:

Attributing a large cluster of traits such as personality, preferences, ability based on a relatively small amount of information about some of the most widely collected demographic attributes -- like gender, age, sex, political views, region etc. 

And notably, for some (most?) examples, the central example is often that judgement is "quick" and based on small pieces of information, such as physical appearance or looks, accent, surname, some official label etc. For example, meeting someone for the first time and either finding out they're group X.

In many cases, the bundle of traits that make up the stereotype are triggered by knowing the demographic category and deployed as assumptions or priors without asking. The priors might come from personal experience, mass media or a combination of them. Another trait pointed out in the wikipedia article is resistance to updating from new information.

Replies from: ChristianKl
comment by ChristianKl · 2021-06-06T18:56:59.471Z · LW(p) · GW(p)

In social psychology, a stereotype is a generalized belief about a particular category of people. It is an expectation that people might have about every person of a particular group. 

By that definition I'm stereotyping a lot when I think that other people who belong to a certain group breath air. I and every other human is stereotyping a lot in that sense. 

It seeems like there are multiple questions that might be interesting compared to the question you asked.

  1. Do you form a belief of about a person that's resistant to updating at a point where you have relatively little information about them?
  2. When you are faced with a person about whom you have little information, to what extend are you willing to have an explicit model of the person. How strongly does that model influence your actions in the context of the person.
  3. How much entropy do you see in the information that's assessible in a few seconds.
Replies from: orthogenesis, orthogenesis
comment by orthogenesis · 2021-06-06T23:45:49.175Z · LW(p) · GW(p)

Though if very few people besides you and I are participating, then I'd concede this conversation is not that fruitful and go discuss something else, I thought your point (2) was very interesting but I feel the ground the term "stereotype" covers is still a bit more narrow than this.

"2. When you are faced with a person about whom you have little information, to what extend are you willing to have an explicit model of the person. How strongly does that model influence your actions in the context of the person."

Stereotypes aren't just any explicit model based on limited information but are particular models based on larger group membership and assumptions of homogeneity in that group membership (for instance, if based on limited information, I look at Jane glancing at me funny and frowning and make up a mental model about Jane that maybe she hates me because some conspiracy against me as evidenced by her frown reminding me of my childhood bully that also frowned against me that way before plotting against me, that's an explicit model of her but not a conventional thing people would call a stereotype -- it's too thoughtfully individualistic, deliberate, explicit and idiosyncratic a model of a person. If I see Jane frown at me, and then think Jane frowns on me because she is a member-of-group and we all know member-of-groups don't like people doing X, which I just did, that falls more into a typical episode of stereotyping based on limited info).

I think a few traits of stereotypes (but not always) involve

  • Being part of the "common knowledge" of a society that most people share (in mass media societies it could be propagated through it, or orally as part of common knowledge elsewhere).
  • Homogeneity assumptions that may have some statistically average grain of truth or sometimes not, but is heavily played up in tropes that people rightly or wrongly see as over-the-top in some settings (people say stereotypes are played up in comedy, ads etc.).  For example, stereotypes of women shopping for shoes are popular and may be backed up by real stats (e.g. data on purchases) but take on a life of their own grander narratives in media.
  • Often passed on without firsthand knowledge to others through hearsay (though firsthand knowledge can confirm or disconfirm it) -- e.g. think of the father telling the young son "what women are like" even before the son has any good mental model from experience, though also heavily backed up by people who have tons of firsthand knowledge insisting it's common sense ("or I know it's stereotypical but it's true, I've been married 40 years"). The thing is stereotypes are things widely recognized by the culture -- something you discovered yourself through years of experience about generalizing about people but are not celebrated in mass culture, common knowledge are generally not called "stereotypes".
  • Not saying all stereotypes have these traits but possessing more of these traits makes something more likely to be called a "stereotype".
  •  
  • 3. How much entropy do you see in the information that's assessible in a few seconds. 
  • Yes, stereotypes seem to be reliant on certain cache'd thought/models and rely on thinking fast. If you think too much about the information and modeling, it almost becomes less  "stereotypical". But not all examples of thinking fast or relying on judgements accessible in a split second are stereotypes though.
     
comment by orthogenesis · 2021-06-06T21:35:38.562Z · LW(p) · GW(p)

I don't know why this is being downvoted so much as I am trying to argue in good faith (I hopefully had tried to give the impression I'm not deliberately stoking culture war but want a clearer rational discussion about this) even though "stereotype" is a fuzzy set. 

Clearly stereotyping is enough of a "thing" that academics (e.g. Lee Jussim etc., people arguing about the validity or invalidity of things like stereotype threat, how important mass media perpetuating stereotypes matter vs. "common sense, people seeing what's in front of them") heavily cite each other and debate intensely over it.

I am thinking about the central examples in social psychology (fast judgement of demographic characteristics that emphasize social differences between groups that are at best statistical, so not "all humans breathe air", particularly traits that are amplified or played up due to their salience in mass culture, society, or media etc like "women like shopping".). 

I hate to do a "I know it when I see it" when it comes to the word stereotyping but even discussing what are the central "traits" of stereotypes are worth discussing.

I do think breaking down those questions makes sense. On your point about mental, explicit models, it seems what most people call "social" stereotypes have certain traits that rationalists would utilize (e.g. being explicit about how you might not be able to predict one individual's traits, but can rely on group averages like "men are more risk-taking") or using correlations of one trait to predict another trait.

Perhaps stereotypes are better framed as a type of prior -- but one that usually apply to people (you rarely hear about animal or inanimate object "stereotypes" but those are usually called generalizations like "the sky is blue -- well in sunny locales", or "dogs are friendly -- mostly") and subgroups or people, whose validity is heavily disputed.

Maybe framed another way, my central question is something like:

Are rationalists more likely to stray from the type of priors based on group-level characteristics that in academia/pop culture etc. are labelled "stereotypes". Because of various reasons -- higher prior on individual variation (less on homogeneity)? 

More focus on analytical thinking -- less gestalt belief in essentialism of clusters even if you use cluster thinking (i.e, the categories are made for the man)? More awareness that even if using stereotypes can be useful, they are a tool only, and stereotypes are in the map not the territory (a poorly drawn but still-occasionally handy map is a trait of the map, the territory is still complex).

Maybe other traits like less reliance on mass media/pop culture (people are rarely explicit on where the "model" for stereotypes come from -- is it personal experience like "stereotypes are really just a name for my personal accumulated experiences, prior updating to posterior, and then many times over and over time I become more accurate so they become second nature" or are their particular sources amplifying stereotypes, like this is true for X% of some group people, but everyone shouts about, jokes about the stereotype as to give the impression that it's 3X% of people,, biasing my priors, amount of updating etc. and if rationalists fall prey to them less)?

Also I'm not trying to use "stereotype less" as "applause light for "yay us, thoughtful individualists" entirely as I agree with the academic Lee Jussim that stereotypes can be useful and academics have overplayed the "things people call stereotypes are inaccurate and useless in providing information" card too much.

I acknowledge there can be stuff that is sacrificed for being less stereotypical and more open-minded (e.g. take longer to think, get mocked as "lacking common sense", "of course group X and Y differ" etc.). But at some benefits like being careful and thoughtful about the (possibly smaller numbers of) individuals who buck the trend or don't follow the stereotype (that society or other individuals might not always see as worth the trade-off vs. quick demographic judgement like in the insurance risk scenario).

Replies from: ChristianKl
comment by ChristianKl · 2021-06-06T22:08:21.432Z · LW(p) · GW(p)

I am thinking about the central examples in social psychology (fast judgement of demographic characteristics that emphasize social differences between groups that are at best statistical, so not "all humans breathe air"

You proposed a definition that's more global. Bailey-and-Motte issues are highly problematic, especially when it comes to politically charged terms. 

You provide no justification for why it would be useful to discuss the issue in those terms instead of using terms that are less politically charged. You provide no explanation about what motivates you to frame the debate this way.

Replies from: orthogenesis
comment by orthogenesis · 2021-06-06T22:48:16.592Z · LW(p) · GW(p)

I don't know a better way of phrasing the central example of category of thinking that is "social generalizations about categories of human that are statistical but commonly discussed in mass media, culture etc." in a clear and concise way other than "stereotypes" or "stereotyping". (One of the most common examples I started off with was fast, physical-appearance-based processing of demographic attributes like sex, appearance, accent, dress etc., which then trigger assumptions about people like personality, dispositions etc. most agree are a component in what people label as "stereotypes").

I was trying to get at a cluster of traits -- the "family resemblance thing" even if I couldn't formalize it well the first time. I was not trying to "Motte-and Bailey" the term "stereotype" but genuinely having a hard time grasping at a less charged term but still had the cluster of traits (in the sense of "what is art" is hard and heavily debated but you could still have a question like do rationalists value art or want to fund art more than average members of the public?).  

By analogy, people talk about westerners thinking more individualistically and easterners prefering collectivism, or STEM-types thinking analytically and humanities thinking more holistically or "people orientation" vs. "thing orientation" all the time, even though all those ways of thinking are politically charged ("what is collectivism" it's a fuzzy set of traits and has a negative connotation in some places but not others) but we can still rationally discuss them.  

I didn't even think "stereotype" was that politically charged in any particular direction (particular stereotypes are, but the concept of there being a thing called "stereotyping" which people of all political persuasions agree exist and people should rationally admit to doing (just like people rationally admit to "generalizing from one example" -- I do myself). I avoided focusing on any one subcategory of stereotyping precisely because I wanted to be more meta (instead of asking "are rationalists more skeptical of gender stereotypes", or "are they more skeptical of national stereotypes"?).

I suppose I could reword to something like "are rationalists less likely to make broad brush social generations from the "general culture" versus firsthand experience and waiting until getting more data firsthand from individuals" or something? Or even "are rationalists more skeptical about broad generalizations about human beings based on limited data about demographic categories than the average member of the public would be?" perhaps chanelling ideas like being aware of "generalizing from one example" or "typical mind fallacy" or "thinking fast and slow". Would that sound better?

And if others don't think this discussion is fruitful (as the initial downvotes show), than that is fine too. I am willing to concede and say maybe I'll spend time on, or discussing, other things. No hard feelings. 

Replies from: ChristianKl
comment by ChristianKl · 2021-06-06T23:55:43.683Z · LW(p) · GW(p)

I was not trying to "Motte-and Bailey" the term "stereotype" but genuinely having a hard time grasping at a less charged term but still had the cluster of traits 

When people engage in bad reasoning because politics mind-killed them it's generally not because they try to reason badly. 

And if others don't think this discussion is fruitful 

It's not just that, it's also potentially costly to have the discussion on LessWrong.

in a clear and concise way other than "stereotypes" or "stereotyping". 

Given that there are bailey-and-motte issues, it's not a clear term. Why do you believe you need a concise way instead of tabooing and explain the cluster that you mean? Having political charged conversations in less concise ways reduces the potential costs of having them.

just like people rationally admit to "generalizing from one example" -- I do myself

It worth noting here that people who form their opinion by "generalizing from one example" instead of by listening to common media are not stereotyping in your classification when they judge people by that generalization. 

Replies from: orthogenesis
comment by orthogenesis · 2021-06-07T00:27:52.800Z · LW(p) · GW(p)

It worth noting here that people who form their opinion by "generalizing from one example" instead of by listening to common media are not stereotyping in your classification when they judge people by that generalization. 

Yes, indeed it would run counter to it. I didn't mean that "generalizing from one example" is "like stereotyping"in that they are similar in what type of reasoning they are, but meant to say they are similar as an example of something people could rationally admit to doing (admit to stereotyping, just like admit to generalizing from one example) and acknowledge the existence of or debate the usefulness of.