Thursday, February 10, 2005

Ethics of Generalization

This raises some tricky ethical issues:
For decades, the science of predicting future criminality has been junk science - the guesswork of psychologists who were wrong twice as often as they were right. But today, the detailed collection of crime statistics is beginning to make it possible to determine which bad guys really will commit new offenses. In 2002, the Commonwealth of Virginia began putting such data to use: the state encourages its judges to sentence nonviolent offenders the way insurance agents write policies, based on a short list of factors with a proven relationship to future risk. If a young, jobless man is convicted of shoplifting, the state is more likely to recommend prison time than when a middle-aged, employed woman commits the same crime.
For [those] who would have stayed clean despite the odds, that heavy sentence would seem unjust. But for states faced with overcrowded prisons and limited budgets, it may not be irrational.

Is it wrong to discriminate in such a way? I've said before that I'm uncomfortable about it: it seems dehumanizing to treat individuals differently purely because of their group affiliations (especially unchosen ones like gender, race, etc.). It denies free will and individual responsibility, suggesting that we are nothing more than the groups that we belong to. That's not the concept of personhood that our society should be promoting.

Another problem is that of ignoring multiplicity, and focusing on just a few salient categories. But a real person (and reality in general) is so much more than that, so one gets a very incomplete picture. I'm a young white male, but I'm also a student, an aspiring philosopher, a liberal, an atheist, and so forth. How many of these would be considered when making a judgment about me? How many should be?

I guess we're just trying to get a more efficient process. If it works, it's not immediately clear that I have any grounds to complain. After all, I'm a utilitarian. Prisons are to protect the innocent, not punish the guilty. If we have good evidence that someone poses no further threat to society, then we shouldn't lock them up. I firmly believe that. But does being a member of a certain group count as such evidence?

It's statistical evidence, for sure - we can say that a random member of group X is more likely to reoffend than a random non-member. But that doesn't really say anything about a given individual (a real person, rather than a mathematical abstraction). I'm not a 'random member', I'm me. The fact that I happen to be a member of group X does nothing to show that I as an individual am any more likely to commit a crime. One can only get into statistical probabilities by abstracting away all the personal details, and dealing with a de-individualized (i.e. dehumanized) number instead. I'm not convinced that treating someone as a number gives you any genuine evidence about that person.

But, returning to efficiency, maybe good personal evidence is not required here. Maybe it's enough for an institution to get good results overall. A good utilitarian could hardly disagree with that. And it's plausible that 'sentencing by the numbers' would get good results, overall. The statistics more or less guarantee it. Viewed over the long run, we are just numbers, not people. Each individual gets averaged out, to become that abstract 'random member of group X' that the formula requires. It would work. So why oppose it?

I think the problems hinted at earlier provide some reason to think that the long-term effects of such discriminatory policy would be harmful. State institutions affect the broader culture. If individuality is no longer a legal reality, it may only be a matter of time until it is no longer a social reality. (When the very legal foundations of society are prejudiced, what sort of message does that send to its members?) Such a cultural disaster would far outweigh the benefits of preventing a few extra crimes.


  1. I've been reading "An introduction to Political Philosophy" by Jonathan Wolff, incidentally, that book you recommended to me a couple of days ago.. and I came across an interesting example from real life of some utilitarian reasoning by Lord Denning. In deciding whether six men, convicted of murder, should be allowed to take civil action against the police (allegedly the police beat them and forced confessions out of them) he said:

    "If the six men fail, it will mean that much time and money will have been expended by many people for no good purpose. If the six men win, it will mean that the police were guilty of perjury, that they were guilty of violence and threats, that the confessions were involuntary and were improperly admitted in evidence and that the convictions were erroneous. That would mean that Home Secretary would either have to recommend they be pardoned or he would have to remit the case to the Court of Appeal. This is such and appalling cista that tevery sensible person in the land would say: It cannot be right these actions should go any further"

    Very, Very nasty, and apparently utilitarian. Or is it? As Wolff says, perhaps such reasoning by the justice system would not increase net happiness or what have you. Perhaps the costs (people would be afraid of being treated as a scapegoat [or, in your example, a number]) would outweigh the benefits (efficiency). I think its simply a matter of adding the indirect cultural costs into the utilitarian equation :). 

    Posted by Patrick

  2. Yes, that's an example of even worse short-sighted utilitarianism. Surely a responsible, accountable justice system is best for society. I agree about the "indirect cultural costs" being key in both problems, and too easily overlooked in utilitarian reasoning... 

    Posted by Richard

  3. > The fact that I happen to be a member of group X does nothing to show that I as an individual am any more likely to commit a crime.

    I think it does - in fact the sum of all the groups to which you belong explains perfectly evewry single on of your actions (give or take a heisenberg uncertainty principle).

    Any conviction must be based on statistical probability if beyond a reasonable doubt is 99% then belonging to a gang and being on drugs and so forth could easily turn 97% into 99.5 % in a rational court. It could also discourage certain behaviours such as drug use or gang membership.
    Now race on the other hand is a bit different. it is a bit like my low iq argument where you might actually be easier on a violent race because you expect slightly less of them ie you create "optimal incentives" just like "i was drunk" might not be an excuse for hitting someone but "i have a mental illness identical to being drunk" might be. 

    Posted by Geniusnz


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.)