Tuesday, September 13, 2005

Racial Profiling

I would like to thank Prof. Thomas for inviting me to respond to his interesting post on profiling and moral progress. He began by contrasting the standard liberal and conservative views on what counts as pernicious racial discrimination. I am afraid that by his classifications, I stand as a poor spokesman for liberals, for I'm wary of affirmative action as well as racial profiling. I take the following anti-prejudice principle to apply universally:

(P) It is morally obnoxious to draw conclusions about an individual person on the basis of their race (or any other involuntary group characteristic for that matter).

I expand on this in my recent post, Why Discrimination is Wrong:
If you treat me differently solely in virtue of my race or gender or age, then you are denying my agency as an individual person. You imply that my group affiliation is sufficiently informative for you to know who I am and what I am like. And this is incredibly insulting to my dignity as an individual. Just because your statistics indicate that young males are disproportionately violent, this doesn't mean that I am violent. If you prejudge me, you deny my individual agency. You deny that I am a distinct person, a free agent, an individual, who cannot be exhaustively specified in terms of my group affiliations. If you prejudge me, you shrink me as a person, down into your little ‘box’. And that’s insulting.

To borrow Thomas’ example, suppose that 90% of car thieves are black. If a robbery occurs and we have no other information, then it seems rational to conclude that the thief is most likely black. There’s nothing racist about this. But I think it becomes morally problematic when we try to apply this knowledge to individuals. If the police have a range of suspects – some black, some not – and choose to put more resources into investigating the black suspects purely on the basis of their race, then I worry that this is racist.

How can I say that? Didn’t I agree that a black person is more likely to have committed the crime? Not exactly. From the fact that the criminal is most likely black, it does not follow that any particular black person is more likely guilty than anyone else. As Thomas notes, it would be absurd to treat every black professor as a potential robber. But what if we don’t know that they are professors? If all we have to go on is their skin colour, then isn’t that some (perhaps weak) evidence that they are more likely to be criminal that we would otherwise expect? From a purely epistemic perspective, this seems hard to deny – assuming the aforementioned correlation between race and robbery. If we abstract away all the other details of the individuals, then we are forced to conclude that an arbitrary black person is more likely to be the criminal than an arbitrary white person.

Have I just contradicted myself? No, you can only conclude that a given black individual is more likely guilty than his white counterpart as a matter of abstract epistemic probability. You know that there are a higher proportion of car thieves in the black population than the white population, so a randomly chosen black person is more likely to be a car thief than a randomly chosen white person. But – and here is the crucial point – this entire chain of reasoning depends upon seeing the individual person as nothing more than a token of a type, a member of a race, when of course each individual is so much more than that – and it’s morally dubious to treat them otherwise.

We reach completely different probabilistic conclusions if we perceive the person as an individual, rather than as a member of a group. The real modal properties of an individual are completely independent of their race. Whether I am black or white, if all my other properties are held constant, then changing my race would have no effect whatsoever on how likely I am to commit theft. The logic of profiling only works in the abstract. When applied to real, concrete individuals, it breaks down, as explained in the post linked above. It’s like an ad hominem fallacy: if you know that someone is unreliable, you might use abstract probabilistic reasoning to judge that their current argument is likely unsound, without even bothering to read the argument itself. But even if there is a sense in which such behaviour may be rational, we also recognize that it is grossly unfair to the argument itself, which ought to be assessed on its own merits. And of course the same is no less true of human individuals.

My objections here leave open the possibility of a morally innocuous form of profiling, i.e. when the profiler takes care not to leap to conclusions about individuals. The most efficient random security checks will sample from those populations that are most likely – as a matter of statistical fact – to contain threats, e.g., perhaps, young males from certain ethnic groups. But as with all such random checks, it is absolutely crucial that the presumption of innocence is maintained throughout. Obviously the vast majority of sampled individuals are going to be innocent, and it would be unfair and unreasonable to harass them under the assumption that they are all “potential criminals”.

Officials can examine a person with the attitude that the person is a criminal. Or, they can do so respectfully and graciously, as a necessity born of unfortunate circumstances.” The former is clearly pernicious, and might not be so easy to avoid as Thomas implies. Since 9/11, certain circles have adopted an irrational fear and suspicion of Arab or Muslim people, wildly out of proportion to the meagre threat that any such random individual poses.

Even if officials manage to maintain greater objectivity, the cultural impact of racial profiling might be much less controllable. As Thomas notes, we have made moral progress over the past few decades: we are now more likely to see each other as fully-fledged individuals, and not mere tokens of a type. We don’t leap to conclusions about individuals simply on the basis of their race. But what effect would profiling have on this trend? Might it simply aggravate the irrational fears of those who see the world in terms of black and white, “us” vs. “them”? (You can imagine the paranoid whispers, “Even the security officers think Muslims are a threat!”)

So yes, we have made moral progress: progress away from abstraction and stereotyping, and towards treating people as individuals rather than prejudging them on the basis of their race. We have made moral progress away from the sort of thinking that underlies racial profiling. But if officials were to elevate the logic of generalization once again, some of this progress would be lost. Granted, if done properly, it might be possible for profiling to be morally innocuous in itself. But what of the unintended consequences? If it could precipitate a moral backslide in the broader culture of our society, is it really worth such a cost?


  1. I wrote a comment yesterday but it didn't go through. Silly Blogger, and silly blogger not copying before posting. Here it goes again.

    As I see it, you have raised three main considerations about profiling based on unchosen characteristics.

    1. Statistical Discrimination. Using racial information can improve predictions and allow for more efficient services (like increased safety at a lower cost).

    2. People Are Not Statisticians. People are unlikely to use racial information properly. Instead, they'll jump to conclusions, fail to treat people with respect, and even provide less efficient services.

    3. Indirect Effects. The long term consequences of discrimination are likely to be bad, including increased prejudice and a tendency to look at people as members of arbitrary groups rather than as individuals.

    I basically agree with you (we were on the same side of this argument at Right Reason), but I think you're being a little too dissmissive of statistical discrimination. The car thief example, for instance, seems to be more similar to profiling for security screening than you acknolwedge. Dedicating more resources to investigating the black suspects can increase efficiency without pre-judging their guilt (the cops can wait to draw conclusions until they finish their investigation).

    Your main concern is with a slippery slope - allowing statistical discrimination could lead to the bad kinds of profiling while legitimizing prejudice. However, taking too firm a stance against discrimination risks creating a backlash led by people who can cite the real benefits of profiling (for our safety, no less). These pro-discrimination subcultures are unlikely to take sufficient care in guarding against the bad aspects of discrimination, should they ever get to put their views into practice. What we have here is another case of our old friend, the precarious pinnacle. I don't know which is more dangerous, the slope or the pinnacle, but it's important to be aware of both risks. So I'm ambivalent about profiling, and open to the possibility of allowing it in important situations when it's based not on individual judgments but on carefully designed institutional procedures.

  2. Blar, good point. I concede that profiling might be justified in some (perhaps rare) situations.


Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)

Note: only a member of this blog may post a comment.