‘Discrimination’ against certain groups supposedly remains a big problem in the modern world. But I have never found a theory that can sensibly explain what this bad ‘discrimination’ is precisely and sensibly distinguishes the sorts of discrimination which are OK from those which are not OK and justifies the difference.
Here’s the challenge: can anyone develop a complete theory of discrimination that makes sense?
Let’s say we know a racial group (or any group) is statistically different on characteristic X. When is it OK to discriminate on that basis if X is something you care about? When, if ever, should we choose to deny ourselves the use of that info? Does it matter what X is as long as you care about it? Does it matter how you get information about these groups and how reliable your information is?
I’m assuming mere errors cannot be justified. The hard question is figuring out when, if ever, using information accurately is a bad thing. We should consider groupings all the way from the fully involuntary (gender/race) through traits that are voluntary to display (sexuality) and ones that are chosen in the usual sense of the word (political opinions, religion, career, obesity).
Hi! I am a young Australian man ostensibly interested in the truth and maximising the total number of desirable experiences that ever occur. My most enjoy reading about the topics listed above. If you share my interests, friend me on 
19 comments
Comments feed for this article
June 4, 2010 at 2:20 am
jaltcoh
“Let’s say we know a racial group (or any group) is statistically different on characteristic X. When is it OK to discriminate on that basis if X is something you care about?”
You shouldn’t discriminate on the basis of race (or gender). You should discriminate on the basis of X, and let the chips fall where they may.
The only exception would be if race/gender itself — not a characteristic that’s merely correlated with race/gender — is essential to the job. This will rarely allow racial discrimination, but one example is if you’re casting a movie/TV show/play in which one character must be of a certain race in order for the drama to make sense. Of course, it’s much more common to require a certain gender to play a certain character. Again, discrimination in this narrow, exceptional case is permissible because the very feature is essential to the job. A mere “correlation” doesn’t justify discrimination, period.
It’s not so obvious that “religion” and “obesity” are “chosen in the usual sense of the word.”
June 30, 2010 at 2:20 pm
Phil Goetz
> You shouldn’t discriminate on the basis of race (or gender).
That’s dodging the question. He’s asking the question precisely because there is no such obvious simple rule. Your rule would forbid all of the following:
- Equal opportunity laws (which discriminate against white males)
- Making it easier for women to get permits to carry weapons (men are 7 times as likely to commit violent crimes as women are)
- Spending more than 0.33% of counterterrorism efforts in the US on Islamic males of middle-eastern descent
- Allowing women to have maternity leave when they’re pregnant
- Scholarships for minorities
- Allowing professional sports teams to hire extremely disproportionately-high percentages of black athletes
- Fraternities and sororities
June 4, 2010 at 2:38 am
Robert Wiblin
“You shouldn’t discriminate on the basis of race (or gender). You should discriminate on the basis of X, and let the chips fall where they may.”
If it’s easy to observe X fully, then nobody would bother using other correlated traits. But what if you can’t (fully) observe X, but you can observe race or gender? Or if it’s cheaper to observe race or gender? That’s surely very often the case.
“It’s not so obvious that “religion” and “obesity” are “chosen in the usual sense of the word.””
I would say they are chosen (in the compatibilist sense) because if you really wanted to change your religion or weight level then you usually could.
June 4, 2010 at 2:53 am
Robert Wiblin
It’s also a bit mysterious why we should crush our instinct for using statistical inference based on the traits we observe given that it on average makes us more accurate and is very useful.
June 4, 2010 at 2:57 am
jaltcoh
“That’s surely very often the case.”
I know. My comment was not premised on the idea that “characteristic X” will always be easy to observe. Since we haven’t defined what X is, of course, it might be difficult to observe.
But that’s just life. Decision making is hard. Trying to decide who has what qualities you desire to perform a certain a job is hard. But people just need to expend *whatever* effort and costs are necessary to make these decisions in a nondiscriminatory way.
If the effort/time/costs start getting out of hand, they may need to settle on an imperfect decision for the sake of finality. You might not be 100% sure that the candidate you’ve chosen for a job has whatever qualities are needed to do the job well (whether it’s intelligence or interpersonal skills or management ability or what-have-you), but you have to do the best you can at deciding these things on their merits. Those who are incapable of making such decisions are in no position to be making hiring decisions at all.
The rule against discrimination must be a moral imperative that transcends short-term economic factors. This is not because there are no incentives cutting the other way. We wouldn’t need a rule if there were no temptation to behave contrary to the rule.
In other words, discrimination shouldn’t be *allowed* because it would be a cheap shortcut. It must be *prohibited* precisely *because* it would, if allowed, provide a cheap shortcut.
As for religion and obesity, they’re more complicated than you’re making them sound. I believe in free will, but people don’t just freely decide in a vacuum: “I’m going to be Catholic and skinny!” “I think I’ll be Buddhist and obese!” Personally, I feel totally free to practice whatever religion I want, or not subscribe to any religion at all. But for many people, the idea of not following your family’s religion doesn’t *feel* like a live option even if you and I can sit around pointing out that it’s theoretically possible.
June 4, 2010 at 3:30 am
Robert Wiblin
For most important traits the reality is ALL we can use is correlated signals. A degree is correlated with competence and intelligence. Presenting well in an interview is also only correlated with doing a good job. Are you going to end the use of degrees and interviews too? And if not, what makes it wrong to infer ability from race, but OK from credentials, body language or apparent ability in interviews? What makes these traits special?
And I’m also not just talking about jobs here. Can I choose my friends based on statistical inference about which ones I will enjoy the company of? My girlfriends? How closely correlated does a piece of evidence have to be before I can use it?
“But people just need to expend *whatever* effort and costs are necessary to make these decisions in a nondiscriminatory way.”
Why should we be willing to pay such possibly extreme costs to avoid errors that might in practice not be very harmful (hiring a person who is only a little worse than someone else)? I just don’t see why we should treat information we garner from these traits or the errors from treating people differently to how we would with perfect information, as special.
“the idea of not following your family’s religion doesn’t *feel* like a live option even if you and I can sit around pointing out that it’s theoretically possible.”
If they wanted to believe something different they could. Perhaps they would be compelled to engage in other practices though. Anyway I don’t care to go on with this unless it’s relevant to the discrimination discussion.
June 4, 2010 at 5:25 am
jaltcoh
I never meant to suggest that there’s generally something wrong with using “correlations.” My only point is that correlations don’t justify race and gender discrimination.
Having a snappy answer to a complex question may be correlated with high intelligence, but it’s obviously not the same thing as high intelligence. I have no problem with an interviewer using this as evidence that a job candidate is smart.
So, again, I don’t object to the use of correlations. I object to the use of correlations *to justify discrimination*.
I’ve talked about job hiring because it’s a convenient example. Yes, it could apply to other things. Choosing a girlfriend is a more complex situation as far as racial discrimination, and gender discrimination is obviously permissible there.
You say you’re worried about the “possibly extreme costs” of *not* discriminating. I don’t see these costs. I did mention “costs,” but I think they’re minor and manageable. I do see the possibly extreme costs of *discriminating*.
You need to look at it not just as a series of narrow thought-experiments about deciding whether to hire person A or person B for one job, but from a much broader societal perspective that includes how the discrimination would realistically play out. In fact, we hardly need to do a thought-experiment about this, since history has so many examples of pervasive race and gender discrimination. How did it play out? Well, to use gender as an example, women have for most of history been barred from most jobs. If you rule out half the population as candidates for most jobs, you’re missing out on an enormous potential for productivity. (Women are just an example; I would say something similar of discrimination against men for certain jobs.)
You might think there’s a more “rational” way to discriminate — but guess what? People are never going to behave purely rationally. People are always imperfect; perfection isn’t an option. We just have to choose among various imperfect systems. And my rough judgment is that a nondiscriminatory system — even if you can find some clever flaw with it — is, on the whole, in the real world, better than a discriminatory system.
And this isn’t just because of the productivity thing, but also because large-scale discrimination in a society infects *everything*, often in ways that are hard to see. There is more to life than just productivity. If, for the sake of argument, a bit of lost productivity is the price we pay for people being able to feel that they’re treated equally and as individuals, I say it’s worth it.
June 4, 2010 at 5:46 am
Louis Krodel
There are many theories of discrimination. These are based on the human brain capacity to form groups. Several experiments have been done to see discrimination at work. One was in a classroom where the teacher only chose blue eyed children. No-one in the class knew this, however, brown eyed children stopped raising their hands. A well known experiment is the: Stanford prison experiment: http://www.prisonexp.org/
June 4, 2010 at 8:37 am
sarkology
The fact that the subject potentially under discrimination is of a particular race/group, should provide evidence for a trait X to the degree that race correlates with X. In theory, (bad) discrimination occurs when you revise your estimate for X based on race beyond what is properly Bayesian.
Usually, when discrimination happens, it’s not merely revising X beyond what is strictly provided by the evidence, but also Ys and Zs and lots more. There is usually also confirmation bias, where knowing race A statistically has trait X, you ignore further information which would revise X unfavorably.
But this is really no use in practice. Since decision processes are supposed to be private. What we can observe are outcomes.
What we could do is model the usual factors that go into a particular decision, holding race/group constant, calculate the correlation of X with belonging to the group. Then predict the outcomes of the decision process, and see if the real world result differs.
This is actually somewhat tractable, if costly.
June 9, 2010 at 1:25 pm
Robert Wiblin
Agree we very often make such errors. That is easy to condemn. But what’s harder is to show that we shouldn’t use info correctly.
June 9, 2010 at 8:08 pm
sarkology
Since we can only deal with discrimination policy-wise on a very crude group level, we should pass anti-discrimination laws when we know that individuals are on average using the info incorrectly. This is simply a consequentialist approach. It does not say anything about whether instances of decision processes are bad on an individual level.
On the individual level, you are commiting discrimination when you are using info incorrectly. If you use it correctly, you are not discriminating.
I am aware this goes against people’s usual idea of discrimination. I’m saying they are wrong. As usual, their morality is based on heuristics, and they stick to those even in cases where they don’t apply. In the case of discrimination, it would be for example, using group outcomes as a proxy indicator of discrimination (on an individual level), regardless of underlying group differences, and not being aware that it’s just a proxy.
June 4, 2010 at 1:31 pm
Robert Wiblin
Jaltcoh: Why the exception for these particular traits (gender/race)? What do you think about inferring info from the other traits I listed?
June 6, 2010 at 6:17 pm
Rita T
It is interesting to read sarkology’s analysis of discrimination avoidance from such a “scientific” point of view. In fact, discrimination has only been recently identified in our history, precisely because we have realised that the “scientific method”, “correlations” and “causality” are not effective ways of explaining human behaviour, and more broadly society. This lead to the birth of critical theory, which has allowed us to explore various social phenomena using different “methods”. For example, the narrative method, or feminist/post-colonial analysis.
Robert, you write “I have never found a theory that can sensibly explain what this bad ‘discrimination’ is precisely and sensibly distinguishes the sorts of discrimination which are OK from those which are not OK and justifies the difference”. The very construction of the phrase implies that ‘discrimination’ is in need of a “theory”, when you can read thousands of modern narratives from people who have and do experience discrimination on a daily basis. So ‘narrative’ analysis tells us that discrimination exists – today. You can conduct a narrative analysis to better understand the nature of discrimination from the point of view of people who claim to experience it to get a better idea.
If we use a critical lens, on the other hand, we can explore the phenomenon of ‘discrimination’ OR determine whether it ‘exists’, by examining power relations between stakeholders and participants in a given setting, social structure or institution. Give me a good example, based on “other traits”, as you say, and we can deconstruct it to see what kind, or indeed, whether, this is ‘discrimination’.
Why do I need an example?? Because if we attempt to determine a “theory” then it needs to be context-specific. This is always the case if you are interested in exploring a social phenomenon, which ‘discrimination’ most definitely is. Basically we need the right tools for the job, ethnographic or critical tools are better suited to this subject than scientific tools, which attempt to rationalise and generalise to the whole population.
June 9, 2010 at 1:20 pm
Robert Wiblin
“In fact, discrimination has only been recently identified in our history, precisely because we have realised that the “scientific method”, “correlations” and “causality” are not effective ways of explaining human behaviour, and more broadly society.”
LOL. I never realised.
I’m not so much asking for a definition of discrimination. Discrimination is when you choose between things. What I want to know is when discrimination is overall bad (from a utilitarian point of view). Obviously we would develop such a general theory by looking at specific cases, judging them, then trying to spot the overall pattern.
June 9, 2010 at 4:16 pm
Rita T
“What I want to know is when discrimination is overall bad (from a utilitarian point of view)”
I would assume that discrimination is ‘bad’ when it results in an unequal distribution of power in a transaction, therefore negating any assumed ‘equal opportunity’. Using the example of a job interview (where people are selected based on explicit characteristics that fit the job):
Person X possesses all the formally desired characteristics for the job but he also belongs to group Y. However the employer has certain stereotypical beliefs about group Y, also hiring a group Y person would run counter to established traditions in her company/institution, traditions which stem from patriarchal, White, colonial belief systems – ideas which have had the most power over the past thousands of years. Therefore the emplyer’s perception of group Y means that person X does not get ‘equal opportunity’ in the transaction taking place. Which has resulted in am imbalance of power between person X and person z (who does not belong to group y). This imbalance of power is a result of discrimination on the part of the employer, whose reasons had nothing to do with the actual job in question. It is the employers hidden judgement of person X that results in discrimination, which could have been avoided if the employer only made overt judgements about person X in relation to the job. These hidden judgements that we all, unfortunately are guilty of at times, can be addressed by applying a lens to our own thought processes, such as a Whiteness lens (in the case of cultural/racial discrimination).
Then we have ‘positive discrimination’, which is aimed at resoting the balance of power in society.
Thus discrimination is ‘bad’ from a utilitarian point of view when it prevents groups of people in broader society from having access to basic things in society (jobs, medical care, etc.).
But now you’ve got me thinking, does the tenet ‘greatest good for the greatest number’ imply that minority groups are collateral damage in the majority’s quest for pleasure and happiness???
Perhaps discrimination is about social justice, and it can’t be looked at from a utilitarian point of view. What say you?
June 9, 2010 at 11:23 am
anonymous
Robert you just got BAMMED by rita T! ouch…..
July 13, 2010 at 2:18 pm
FatMan
Short Answer: The individual is not the group.
Long Answer:
Group data is based on rough statistical average and it’s just plain lazy thinking to overlook individual characteristics in favor or presuming characteristics based on group affiliation.
Not just that but by assigning the statistical average to someone automatically without concern for their individual abilities seriously inhibits them from participating in social and economic life. A black man with a 150 IQ may be doomed to a life of washing dishes if society decides to treat him only as a statistically average black male.
The positive utility you gain from intellectual laziness is far outweighed by the harm caused in miscategorizing others.
And there’s that whole oppression thing that tends to occur when it becomes socially acceptable to treat a large group of people as inferior (for whatever reason).
And nothing I said above would prevent useful discrimination (like girlfriends, or degrees, etc.).
September 11, 2011 at 3:14 am
Il guaio delle lotte anti-discriminazione… « fahreunblog
[...] Wiblin lo dice meglio: …Discrimination’ against certain groups supposedly remains a big problem in the modern world. [...]
January 4, 2012 at 2:47 pm
Febri Asmara
“no law” about discrimination.. :’(