When (and Why) Is Discrimination Acceptable?
Information is useful, but sometimes people insist on ignoring it.
Posted Jun 04, 2013
As a means of humble-bragging, I like to tell people that I have been rejected from many prestigious universities; the University of Pennsylvania, Harvard, and Yale are all on that list. Also on that list happens to be the University of New Mexico, home of one Geoffrey Miller. Very recently, Dr. Miller has found himself in a little bit of moral hot water from what seems to be an ill-conceived tweet. It reads as follows:
“Dear obese PhD applicants: if you don’t have enough willpower to stop eating carbs, you won’t have the willpower to do a dissertation #truth“.
Miller subsequently deleted the tweet and apologized for it in two follow up tweets. Now, as I mentioned, I’ve been previously rejected from Miller’s lab – on more than one occasion, mind you (I forgot if it was 3 or 4 times now) – so clearly, I was discriminated against. Indeed, discrimination policies are vital to anyone, university or otherwise, with open positions to fill. When you have 10 slots open and you get approximately 750 applications, you need some way of discriminating between them (and whatever method you use will disappoint approximately 740 of them). Evidently, being obese is one characteristic that people found to be morally unacceptable to even jokingly suggest you were discriminating on the basis of. This raises the question of why?
Well, not so fast. Let’s say, for the sake of this argument, that obesity was actually a predictor of graduate school performance. I don’t know if there’s actually any predictive value there, but let’s assume there is and, just for the sake of this example, let’s assume that being obese was indicative of doing slightly worse at school, like Geoffrey suggested; why it might have that effect is, for the moment, of no importance. So, given that obesity could, to some extent, predict graduate school performance, should schools be morally allowed to use it in order to discriminate between potential applicants?
I happen to think the matter is not nearly so simple as predictive value. For starters, there doesn’t seem to be any widely-agreed upon rule as for precisely how predictive some variable needs to be before its use is deemed morally acceptable. If obesity could, controlling for all other variables, predict an additional 1% of the variance graduate performance, should applications start including boxes for height and weight? While 1% might not seem like a lot, if you could give yourself a 1% better chance at succeeding at some task for free (landing a promotion, getting hired, avoiding being struck by a car or, in this case, admitting a productive student), it seems like almost everyone would be interested in doing so; ignoring or avoiding useful information would be a very curious route to opt for, as it only ensures that, on the whole, you make a worse decision than if you hadn’t considered it. One could play around with the numbers to try and find some threshold of acceptability, if they were so inclined (i.e. what if it could predict 10%, or only 0.1%), to help drive the point home. In any case, there are a number of different factors which could predict graduate school performance in different respects: previous GPAs, letters of recommendation, other reasoning tasks, previous work experience, and so on. However, to the best of my knowledge, no one is arguing that it would be immoral to only use any of them other than the best predictor (or the top X number of predictors, or the second best if you aren’t using the first, and so on). The core of the issue seems to center on obesity, rather than discriminant validity per se.
The twist was that these classifications of high- and low-risk either happened to correlate along racial lines, or they did not, despite their being no a priori interest in discriminating against any one race. When faced with this situation, something interesting happens: compared to conservatives and moderates, when confronted with data suggesting black people tended to live in the high-risk areas, liberals tended to advocate for disallowing the use of the data to make profit-maximizing economic choices. However, this effect was not present when the people being discriminated against in the high-risk area happened to be white.
In other words, people don’t seem to have an issue with the idea of using useful data to discriminate amongst groups of people itself, but if that discrimination ended up affecting the “wrong” group, it can be deemed morally problematic. As Tetlock et al (2000) argued, people are viewing certain types of discrimination not as “tricky statistical issues” but rather as moral ones. The parallels to our initial example are apparent: even if discriminating on the basis of obesity could provide us with useful information, the act itself is not morally acceptable in some circles. Why people might view discrimination against obese people morally offensive itself is a separate matter. After all, as previously mentioned, people tend to have no moral problems with tests like GRE that discriminate not on weight, but other characteristics, such as working memory, information processing speeds, and a number of other difficult to change factors. Unfortunately, people tend to not have much in the way of conscious insight into how their moral judgments are arrived at and what variables they make use of (Hauser et al, 2007), so we can’t just ask people about their judgments and expect compelling answers.
Though I have no data bearing on the subject, I can make some educated guesses as to why obesity might have moral protection: first, and perhaps most obvious, is that people with the moral qualms about discrimination along the weight dimension might themselves tend to be fat or obese and would prefer to not have that count against them. In much the say way, I’m fairly confident that we could expect people who scored low on tests like the GRE to downplay their validity as a measure and suggest that schools really ought to be looking at other factors to determine admission criteria. Relatedly, one might also have people they consider to be their friends or family members who are obese, so they adopt moral stances against discrimination that would ultimately harm their social ingroup. If such groups become prominent enough, siding against them would become progressively costlier. Adopting a moral rule disallowing discrimination on the basis of weight can spread in those cases, even if enforcing that rule is personally costly, on account of not adopting the rule can end up being an even greater cost (as evidenced by Geoffrey currently being hit with a wave of moral condemnation for his remarks).
References: Hauser, M., Cushman, F., Young, L., Kang-Xing Jin, R., & Mikhail, J. (2007). A dissociation between moral judgments and justifications. Mind & Language, 22, 1-21.
Tetlock, P., Kristel, O., Elson, S., Green, M., & Lerner, J. (2000). The psychology of the unthinkable: Taboo trade-offs, forbidden base rates, and heretical counterfactuals. Journal of Personality and Social Psychology, 78 (5), 853-870 DOI: 10.1037//0022-3518.104.22.1683