Insightful comments from J.C. Oleson about using empirical research on risk to reoffend at sentencing

One of the most enriching aspects of this blogging gig for me has been the opportunity to correspond with some brilliant people who are also motivated to do good. I count J.C. Oleson among those folks.  Last evening I received an e-mail from Dr. Oleson. I was struck by how thoughtful it was, and asked if I could post it. He gave me permission to do so. Here it is:

Dear Judge Kopf,

Thank you for your email, and for the link. Your postings have been interesting, thoughtful, and provocative, and the comments have been terrific. I liked Peter H’s comment about the 1866 Civil Rights Act, and have thought a bit about how far the prohibition extends. Obviously, you cannot have one set of penalties for one racial group and another set for another racial group, but I wonder if it would really be an obstacle if race was part of a multi-factor risk assessment. The whole point of my article “Risk in Sentencing” was to suggest that actuarial approaches are so attractive that EVEN RACE might be permissible. And if race can be permitted in, then everything else follows. I don’t honestly think that the Supreme Court would – or should – condone sentencing disparities based on race data – but the point is that it COULD (holding community safety to be a compelling state interest). Of course, even if “race” is NOT permitted in, and even if gender/sex is not permitted in, many of our standard considerations (e.g., education, work history, socioeconomic class, community ties) that correlate with race and gender might very well be permitted. Once you really begin to scrutinize the idea of a sentencing “fact” and to consider the linkages between the fact and the sentencing decision, it becomes epistemologically very complicated. I think that’s what I like about your postings and the comments – there seems to be a recognition that this is a problem that is already with us, and affords no simple solutions.

· These variables highlight an incommensurable tension between utilitarian and deontological bases of punishment – being poor may exculpate Jean Valjean under just deserts but it also predicts greater risk of offending and recidivism – federal judges, simultanenously trying to punish both past and future crimes under 3553(a) are left without a compass

· People dislike the “sentence-o-matic 1000” but rejecting algorithms and automation does not prevent assessments of risk – it just means that the human estimates are likely to be more idiosyncratic and less accurate

· People are more receptive to using these variables in mitigation, but this may be a case of wanting one’s cake and eating it too – if the variables can exculpate, how is it that they do not also demonstrate blame?

· People are more receptive to using risk instruments at the back end of sentencing (parole, probation conditions) but why should we be more comfortable with that than with sentencing? Many offenders would prefer prison to highly restrictive intensive supervision probation – it’s not a case that sentencing is “serious” while community corrections decisions are inconsequential

The problem of whether to use this data (how much, and in what form) is already here. The AG’s remarks and the work on the MPC provisions underscore the importance of the question. It’s wonderful to see this discussion on your blog (which I have now bookmarked) and it may be a question for the Judicial Conference and the Sentencing Commission to study.

With very best wishes,


J.C. Oleson, Senior Lecturer in Criminology
Director Research, School of Social Sciences
Book Review Editor, Australian & New Zealand Journal of Criminology
Sociology, University of Auckland

 So, dear readers, what do you think?


14 responses

  1. “People are more receptive to using these variables in mitigation, but this may be a case of wanting one’s cake and eating it too – if the variables can exculpate, how is it that they do not also demonstrate blame?”

    The word “blame” seems to give away the game here, no? If our goal is a utilitarian one, then blame doesn’t enter into it. And the troubling point is that the very factors that mitigate a defendant’s blamefulness are the ones that predict recidivism — the result of actuarial sentencing would in many cases be to punish the less blameworthy defendants more harshly, precisely because they are less blameworthy.

    I think Prof. Oleson is correct that we have an intractable philosophical dilemma here between utilitarianism and deontology. But here’s the real rub: our penal system is deontological. It just is. Our prisons are awful, awful places. A shift to a utilitarian theory of sentencing would be morally indefensible without a simultaneous shift away from the extremely punitive nature of incarceration. If the point of long sentences is pure incapacitation, and we conclude that the people who most need to be incapacitated are also the people least to blame for their actions, there is simply no justification for treating people so poorly while they’re in the clutches of the justice system.

  2. Griff,

    Your point that if we shift to a ulitarian approach we must make prisons better is one I agree with entirely. Sadly, I think that is true even if you don’t shift to the ulitarian approach. But, given the fact that is all about saving money, I doubt very much you will see your hope fulfilled.

    All the best.


  3. Mike,

    No. The point is to lock up all the poor and violent people and then treat them well. (Tongue partly in jowl.)

    All the best.


  4. That wasn’t really my point, although my actual point was made in too roundabout a way. What I meant was that it would be immoral (bordering on monstrous) to shift to a utilitarian approach, because there is no reasonable probability that our penal system will start treating convicts better.

  5. Pingback: Better to be Sentenced by Ostrich or Pigeon? (Update) | Simple Justice

  6. Judge:
    Consider this one sentence: “I don’t honestly think that the Supreme Court would – or should – condone sentencing disparities based on race data – but the point is that it COULD (holding community safety to be a compelling state interest).” With due respect to Dr. Oleson, I don’t quite understand what this means. After all, the fact that I COULD win the lottery in no way increases my bank account. More to the point, the admitted attractions of actuarial/statistical sentencing models should not blind us to the very real danger of using immutable characteristics in same. If justice is to mean anything it must mean that punishment can only be imposed upon that which was within the voluntary choice of the offender, i.e., that which is rightfully subject to moral and legal judgement. This runs directly counter to the consideration of an an immutable characteristic–either alone or in concert with other, more acceptable, factors–in punishing criminal defendants.

  7. Robert,

    Here is what I would argue: “Race,” in the sense that it may be correlated with crime, may be taken into account if the government has a compelling governmental interest such as incapacitating dangerous offenders. However, black people, brown people, yellow people and white people are not predisposed to the commission of crime simply because of their skin color or their self identification. In this sense, race becomes a proxy for various cultural phenomena.

    Thinly sliced, you say. Yes, indeed, I reply.

    All the best.


  8. RGK,

    Your argument seems to suggest a violation of the Civil Rights Act of 1866.

    That any person who, under color of any law, statute, ordinance, regulation, or custom, shall subject, or cause to be subjected, any inhabitant of any State or Territory to … different punishment, pains, or penalties… by reason of his color or race, than is prescribed for the punishment of white persons, shall be deemed guilty of a misdemeanor, and, on conviction, shall be punished by fine not exceeding one thousand dollars, or imprisonment not exceeding one year, or both, in the discretion of the court.

    Best Regards,

  9. Judge Kopf:

    I’ve read the various posts and comments with interest and, I must admit, a bit of a headache because this is a difficult subject to think through. That said, I read your reply (above) to Robert, and I have a couple of questions. If we agree that people of various races are not predisposed to criminality because of their races, why should we ever use race even as “a proxy for various cultural phenomena”? Assuming (as I do) that there’s not a one-for-one correlation between race and the cultural phenomena, why not just look at the cultural phenomena, thereby being more precise and avoiding the risk of overinclusion?

    Off to get Tylenol.


  10. Dr. Oleson,

    Thanks for commenting on my mention of the Civil Rights Act of 1866. I remain however unconvinced that, in the context of that Act, and the context of 28 U.S.C. § 994(d) that there is a statutory case for use of race, even as part of an actuarial approach. Congress has made plain on multiple occasions that race is not acceptable as a factor for variation in sentencing. This may be Congress being deontological, but if it is, so be it.

    Congress sets the terms of sentencing (as long as they accord with the Constitution), and Congress’ intent is unambiguous: Sentencing should be “entirely neutral as to the race … of offenders.” 28 USC 994(d).

    Further, the Civil Rights Act of 1866 is worded extremely broadly. “[S]uch citizens, of every race and color… shall be subject to like punishment, pains, and penalties, and to none other, any law, statute, ordinance, regulation, or custom, to the contrary notwithstanding.” Congress says that there shall be like punishment and none other. This isn’t a strict scrutiny question. This is statutory construction. No compelling interest overrides a clear congressional prohibition.

    To the ethical point about consequentialism versus deontology, I think you are correct that this an appropriate ethical debate between the two. I would also argue that Scott Greenfield on his blog has been making a good argument for moral particularism, especially in this most recent post.

    But I think that importing the risk based models you propose into current sentencing law, would be deeply unconsequentialist.

    If we are to be act consequentialists in sentencing, then the moral framework must be taken as a whole. That is, there can be no element of just deserts; the sentence is wholly forward looking. This is not what would happen. Rather, we would get the higher of the two of the just deserts deontic sentence, or the risk-assessment based consequentialist sentence. This is a moral mish-mash, which squeezes out both fairness and benefit to society.

  11. David,

    Great question. The anwer to your question is this: We use race as a proxy for cultural phenomena for two reasons: (1) we are talking about a big basket of factors and (2) it would be almost imposssible to tease out separately each one.

    All the best.


  12. Judge —

    I can understand that the basket of cultural phenomena might be very large indeed and that it might be difficult to tease out one from another. However, I don’t quite follow why either of those is a justification for simply using race as a proxy. If there are so many factors and the task of teasing them out would be Herculean, shouldn’t that lessen our confidence in using race as a proxy because we don’t know just which collection of cultural phenomena constitute the proxy we are designating? Put another way, if we accept (as my original question did) that people are not predisposed to future criminality or violence simply because of their race and that there is no one-for-one correlation between race and future conduct, how could we be comfortable setting a term of imprisonment based so plainly on a tool of effiency rather than on one of accuracy? (I realize that the last clause of that question will spawn all sorts of new questions, including “what is the alternative since there is no perfect way to predict future conduct and such predictions are necessary to depopulating the prisons when possible?)

    Perhaps we could avoid this if we were simply to amend Article III to require presidents to appoint soothsayers as district judges. It might be easier.

    All best,


  13. Of course, the whole endeavor is somewhat complicated by the fact that the “basket of cultural phenomena” at issue includes centuries of slavery, white supremacy, segregation, Jim Crow, redlining, etc., and is perpetuated in the present day by systemic racism and implicit bias. So, query whether the best solution is to lock people up longer based on their race, or even based on the “cultural phenomena” that correlate with their race.

%d bloggers like this: