Friday, January 13, 2012
Having recently begun reading Peter Singer’s classic text Practical Ethics, I was interested to see his stance on Affirmative Action. I don’t completely agree with Singer’s final conclusions but his lines of reasoning have nonetheless been extremely illuminating - and it allows me to discuss something else that’s been on my mind lately, which is our use of proxies in decision-making.
Singer broadly subscribes to the ethical system of Preference Utilitarianism. I’ve touched on Utilitarianism before and I don’t really have the space to delve into its complexities, but the basic premise is that we should judge a specific action as ethically good if it promotes happiness in sentient beings, and judge it as ethically bad if it promotes suffering, rather than judging them on how well they fit rigid, generalised laws (such as “Thou shalt not X”). Preference Utilitarianism differs slightly in that it judges actions as good if they advance a sentient being’s preferences, not its happiness. The distinction is pertinent but very subtle.
It deserves a blog post of its own to discuss the finer points, but speaking broadly, I too consider Utilitarianism the most useful moral schema yet devised.
In discussing affirmative action, my view has always been roughly this: the idea that a person who is better-qualified for a position should be denied it to fill a quota is a false economy. It may buck historical trends of discrimination, but it is discrimination nonetheless. My view is that while we absolutely should be acting to eliminate discriminatory hiring practises, this is not really with the goal of having more black people as doctors, or more women as CEOs - it is with the goal of having it so that a person need not worry about their race or gender, and be assured that if they are smart, talented and hard-working enough, they will get the job they deserve. Having more representative demography is a side-effect, not the goal.
Historically, humanity tends to avenge itself from one extreme by flying to the other - when a group is oppressed for a long time and eventually throws off their shackles, they tend to end up taking over and being just as oppressive as the last lot were. I am therefore very, very wary of any system that tries to compensate for disadvantage by giving disproportionate advantage - I would much rather have slow progress that resulted in equality than fast progress that quickly swung around to a different kind of inequality.
In my view, if there happen to be fewer women in a certain line of work - solely because they choose not to enter it - then we should respect that choice. There will never be complete demographic representation across all careers due to various cultural factors (and to some extent, statistical trends of inherent ability) and I think that’s fine. This isn’t to say that the current inequalities are due to such factors - they aren’t. But it is to say that a person’s skin colour is the wrong metric, and that the key is to judge people individually. While it is right to say “We should have hired Jim, he was more qualified than the white guy” it is wrong to say “We should have hired Jim because we need to hire more black guys.” More on this later.
In any case, Singer points out that under his system we must maintain equal consideration of interests - that is, your goals are no more or less important than another person’s goals, even if that person is of a different race, gender, sexuality etc, or if they are a relative or personal friend of mine. This is a very useful system for making a lot of decisions but in certain cases it proves kind of useless.
Okay, so you have two people, and both want the job. You mustn’t favour one’s interests over the other…so how, then, do you resolve the conflict? You have two applicants and only one job - you must use some heuristic to decide. We’ve covered that a person’s gender or race is not an acceptable means of differentiating between them, so what is acceptable? The short version is, due to Utilitarian gains that would result from how well they would perform the job, you must hire the person who will do the job best. Thus in such situations I have always taken “Who is better qualified?” as a proxy-question for “Who will promote the most happiness?”
But Singer uses the case study of Regents of the University of California vs Bakke, in which Bakke sued UC Davis for introducing a quota system to their medical program, so that Bakke missed out on a spot despite having better scores on the entrance exam than some who made it in. Singer points out that Bakke’s argument is essentially based on the premise that as the smarter student he had a right to the position in the course - but that under Singer’s model, he and whoever took his place have exactly equal rights because they have equal consideration of interests. My initial response was as above - the more qualified kid should get it. However, Singer raises some interesting points.
The main reason a black kid might struggle to get into such a course is not because of the inherent intelligence of black people, but is because of the specific socio-economic circumstances of his upbringing. The kid may be as smart as or smarter than Bakke, but since he didn’t have the same advantages growing up, his test scores are stunted. Therefore, the most qualified kid - in a more abstract sense - may not be reflected in the test scores, and we should take this fact into consideration when filling spots in a medical course.
Here we have taken MCAT scores as a proxy for “How good a doctor they will be.” Singer points out that this is wrong, and that if our goal is truly to produce the best doctors, then we should have no problem with rejecting a student whose test scores are artificially inflated due to affluence, in favour of a student whose test scores are artificially lowered due to poverty; from there he concludes that it is therefore just for colleges to practise affirmative action in their admission principles.
Singer hasn’t gotten it quite right, though, and he’s making the same mistake I did. Okay, so a person with a worse socio-economic background is likely to have been academically disadvantaged, and their lower MCAT scores may not reflect their true career potential. Notice how I didn’t mention race or gender anywhere in there? Yes, skin colour is strongly correlated with socio-economic status, but it is not a direct causal relationship.
Since socio-economic status is the thing you are looking for, that must be the metric on which your adjustments are based - not skin colour. There are, after all, poor white people who are disadvantaged academically, and rich black people who are not. Even socio-economic status is not a foolproof indicator - a poor family who encourages their children, reads with them and helps with homework is likely to produce students with better test scores than a rich family who does not value education and spends no time with their children. But in the meantime, while we search for a better metric than socio-economic status, we must use the best metric we have available - not the proxy metrics of race or gender.
You need to do some research, of course. See if there actually is a correlation whereby students from a poor socio-economic background improve their class ranking from MCATs to final exams, or have lower-than-expected drop-out rates, or finish on time more often, or have better careers after leaving college. I suspect that they might. If they do - that’s your metric, right there. Don’t even ask about their skin colour, just ask for their parents’ income bracket.
It’s also worth noting that even this is a fairly poor way of doing things since a poor smart kid might be so stunted as to fall below the threshold of visibility - that is, they might not even take the MCAT, or indeed might not survive/remain outside prison long enough to do so. Especially given every other problem it causes, fixing the initial conditions - opportunity inequality arising from wealth inequality - is a much higher priority, but given that tweaking admissions is a much easier task, it’s quite a good interim measure.
The real take-away from all this, though, is not a lesson about affirmative action - a more general lesson can be learned. While it is often necessary to use proxies - something that stands in the place of what you actually want - it is important to remember that they are just proxies, and what those proxies stand for, so that you can discard them in the situations where they don’t fit as well.
This applies to everything. We don’t want to take licenses from the elderly because their independence is important, but that independence is only a proxy for the happiness it is likely to provide. At a certain point the danger to themselves and others outweighs the advantages gained by independence, and we must discard the proxy.
(Also, note how I used “elderly” as a proxy for the things we should actually test, such as visual acuity, reaction times and motor skills)
We don’t want to needlessly restrict people’s choices to take certain drugs, but that freedom of choice is only valuable in so far as it promotes happiness in that individual. At a point where it becomes clear that they are doing more damage to themselves by abusing that drug than they are gaining in joy from taking it - and simply assuming that when this point is reached they will choose to stop is untenable given what we know about addiction - we must discard the proxy and save them from themselves, be it by intervention, regulation or outright banning.
These are just the first examples that come to mind. The point is, though, it takes a great deal of thinking to peel back the layers of justification to the core reasons why we value things. But it is time well-spent, since often our pursuit of sub-goals can actively damage our true, long-term goals.