« A solution for the wrong problem | Main | The best employment outcomes for law students »

Friday, May 04, 2018

Reply to Galoob's "Kolber's Teaser"

In PMR, I argued that retributivists need to believe at least nine propositions in order to inflict just punishment on a particular offender. Reasonable retributivists, I claimed, will have some doubts about each proposition, and they should multiply levels of confidence in each proposition to determine confidence in the nine-proposition conjunction.

In Kolber's Teaser, Stephen Galoob presents two main challenges. First, he argues that the nine retributivist propositions I address are not independent of each other, hence we cannot multiply probabilities to determine confidence in the conjunction. (Larry Solum, as Galoob mentions, made a similar critique.) Galoob notes, for example, that if one proposition is "The Patriots win the Super Bowl this year," and another is that "Patriots quarterback Tom Brady is named Super Bowl MVP this year," we cannot simply take the probability of each and multiply them. The reason is that the probability of these events are dependent on each other. If the Patriots win the Super Bowl, the chances Brady is named MVP increase substantially.

The short answer is that I don't make the mistake Galoob identifies. I ask for confidence in proposition #1, but when I ask about the next matter, I ask for the reader's confidence in the matter given the truth of proposition #1. For the third matter, I ask for the reader's confidence in that matter given the truth of #1 and #2, and so on for all nine propositions. Here's the math on conditional probability that makes that appropriate: "If events A and B are not independent, then . . . the probability that both events occur . . . is defined by P(A and B) = P(A)P(B|A) [where P(B|A) means the probability of B given A]." Back to football, there would be no error in saying that the probability that both the Patriots win and Brady is named Superbowl MVP is calculated by multiplying the probability the Patriots win the Super Bowl times the probability that "Tom Brady is named MVP given that the Patriots win the Super Bowl."

So the short answer is that there is no mathematical error. I'm sensitive in the paper, though, to the concern that underlies both Galoob and Solum's commentaries. Here's how I would put it: "Ok, Kolber, you can satisfy the mathematicians this way. But can we really assess the probability of, say, "wrongdoing warrants suffering given that we have the sort of free will required for moral responsibility"? Admittedly, this is challenging. There's no question that the probabilities I ask for are hard to assess, and they are probably made harder when you have to do so in a conditional way. But: (1) The substance of the matters underlying the propositions are clearly somewhat independent. You can believe we have free will but not think that wrongdoing warrants suffering. The independence of many of the other underlying matters are even clearer than for these two; and  (2) The task here is not addressed to laypeople but to retributivist theorists. They claim to have a justification of punishment. It seems reasonable to ask theorists who claim to justify punishment of individual offenders how confident they are in that claim. If they can't answer the question, at least in some approximate sort of way, then that's a serious demerit of retributivism. 

I respond to Galoob's second point after the jump.

Galoob writes (footnotes omitted):

The retributivist could respond to Kolber’s teaser gambit by arguing from a normative division of labor. Such an argument might invoke John Rawls’s distinction between “justifying a practice” and “justifying a particular action falling within” a practice. The former task implicates different normative principles and identifies different types of considerations as relevant than the latter does.

Assume that a retributivist accepts the normative division of labor argument with regard to the justification for punishment. One implication of this argument is that uncertainty related to the justification of a practice is analytically distinct from uncertainty related to the justification of actions within a practice. . . .

The normative division of labor argument in response to Kolber’s gambit, then, takes two steps. First, different normative principles apply to background- and case-level propositions, and categorically different normative and empirical considerations are relevant to establishing the truth-value of these propositions. Yet, Kolber’s gambit only works if the same (or commensurable) principles and considerations apply to the evaluation of background- and case-level propositions. Second, in light of discrepancies between background- and case-level propositions, uncertainty about one category of proposition seems incommensurable (or, at least, not easily comparable) with uncertainty about the other. Yet, Kolber’s gambit assumes that uncertainty about one type of proposition is comparable with uncertainty about the other. Therefore, according to this argument, Kolber’s gambit is mistaken.

My main response is that in order to justify punishment of some particular offender, retributivists have to simultaneously believe all nine propositions. They can't simply wall off some of them. They can't say, "Well, I believe this offender committed all the elements of the offense and that his sentence is proportional," but I don't have to tell you the probability that anyone ever has free will because that concerns a different question that applies to all offenders and not just this one. This is a bit like the flip side of Galoob's first criticism. Just as I say you have to consider each proposition given prior propositions, retributivists can't pick and choose propositions and purport to have justified the full conjunction. 

I think that settles the matter, but Galoob makes another point. In the paper, I criticize reliance on Hart's separate questions approach. Galoob anticipates that I will offer this criticism as a response to his own regarding hypothetical retributivists who refuse to multiply all nine propositions. Galoob claims that, even if my criticism of the separate questions approach is correct, many retributivists in fact subscribe to that approach or something like it. Moreover, since I purport to criticize retributivism from a perspective internal to retributivism--a perspective that even retributivists could accept--my denial of the separate questions approach would be external to retributivism. If so, he argues, I can no longer claim to give an internal criticism of retributivism. Very interesting!

I would again refer to my answer two paragraphs up. If a doctor is asked the chances that a patient has both diabetes and heart disease and therefore requires some particular drug, the doctor can't passively decline to answer on the artificial grounds that these are separate questions. We simply need to know the conjunction to take action (whether it's to give some particular drug or to actually punish real human beings). The same obligation to consider all kinds of uncertainty applies even when we mix moral and descriptive forms of uncertainty (e.g., should the doctor perform the surgery when there's both medical uncertainty about whether the surgery will succeed and moral uncertainty about whether, even if it did, it's the right decision given the patient's expected quality of life?). So I think that if the view Galoob describes is internal to retributivism (namely, that uncertainty can be bracketed off by focusing on different questions at different stages of inquiry), retributivism may have more serious problems than my paper adverts to.

At the same time though, I don't see the view Galoob raises as internal to retributivism. It's true that one can distinguish questions that apply to all offenders and those that apply to particular offenders. Indeed, I do so simply as a matter of organizing my paper. But the connection between separating those questions and the questions Rawls or Hart focused on (e.g., the justification of institutions of punishment vs. the justification of the amount of punishment distributed to some particular offender) is weak. They were not speaking of our confidence in our individual beliefs and how they affect our overall conclusions. So if there were actually a substantial number of retributivists separating out the particular questions Galoob references, then my critique might indeed be at least partly external to retributivism. I don't think enough retributivists have even addressed the matter in the context of belief under uncertainty for us to say that it's internal to retributivism. But as I think the doctor example shows, retributivists should not (and probably do not) find that sort of reasoning appealing. 

Galoob concludes with the following very generous sentiment: 

Professor Kolber is one of the most astute critics of retributivism and one of the most innovative contemporary punishment theorists. Has he succeeded in confounding retributivism, of knocking it from its spot as the preeminent contemporary theory of punishment? Perhaps not this time. But I would not bet against him.

Let me add that I am familiar with Galoob's work as well and echo his compliments right back at him. I wouldn't bet against him either this time or any time, and I thank him for his very thoughtful commentary. 

Posted by Adam Kolber on May 4, 2018 at 11:25 AM | Permalink


Thanks for coming by to comment, Stephen!

"If we assume the truth of Proposition (4), we are saying something important about the truth of Proposition (7)." Agreed. Proposition 7 asks about police and judicial conduct in a particular case, and as you note, it probably *cannot* be just if all state coercion is unjust. That means when you answer (7), your probability should be higher than it would have been had you not assumed the truth of (4). Yes, the underlying issues are related, and yes, the assumptions about prior propositions will affect probability estimates. We assume the truth of the prior propositions precisely to avoid the criticism that you and Larry make. If we didn't assume the truth of the prior propositions, then your criticism would be spot on, and the propositions would be hopelessly dependent on each other.

I think it would be great if retributivists want to try to spell out the argument you outline in your second comment. If they can address the sorts of concerns I raise above (e.g., you can't subscribe to a theory that requires simultaneous belief in a set of propositions but only consider one's uncertainty about subsets of the propositions), I will be all the more impressed. Thanks again!!!

Posted by: Adam Kolber | May 4, 2018 4:21:43 PM

On the more substantive point, you state that “I don't think enough retributivists have even addressed the matter in the context of belief under uncertainty for us to say that it's internal to retributivism.”

I think you’re absolutely correct here. Retributivists have, by and large, ignored or deemphasized questions of criminal procedure and political philosophy, questions that your approach forces them to confront. However, just because most have ignored this question doesn’t mean that all have—Doug Husak and Mitch Berman come to mind as retributivists who have addressed it, albeit largely obliquely.

Moreover, even if retributivists have not addressed the question, it is still possible to formulate a range of possible responses that retributivists might make. The Division of Labor argument I outline in the piece seems like a plausible response. It would emphasize the significance of background propositions (especially those related to free will and moral responsibility) and use these to inform case-level propositions. Indeed, something like this Division of Labor response might be read out of Husak and Berman’s positions.

Your paper, of course, seems to rule out the Division of Labor argument based on your construal of what a proper answer to the questions you’re posing might look like. My point here is that your construal of the question rules out, in advance, a likely route for retributivists to respond. This strikes me as dialectically unfair, a way of steering retributivists toward hybrid theories based on little more than methodological assumptions.

Now, the best theory of punishment may indeed be a hybrid theory (and, moreover, a hybrid theory of the consequentialist-flavored variety that you describe). But that seems like where we should end up at the end of the debate, rather than at the beginning.

Posted by: Stephen Galoob | May 4, 2018 3:12:41 PM


I agree with many of your paper’s big points, but I have some quibbles with the smaller stuff.

Regarding the methodological criticism, you state that “The short answer is that I don't make the mistake Galoob identifies.”

My main criticism here was that your propositions have linked truth values—in many cases, the truth of one proposition affects (and, in some cases, determines) the truth of another. If so, then you shouldn’t treat the propositions as independent from each other for the purposes of determining the retributivist’s burden. Moreover, you can't assume the truth of an earlier proposition in assessing the truth of a later proposition, because that assumption changes the truth-value of the later proposition. I don’t think that your remarks here do anything to change that methodological worry.

To take an example I discuss in the paper, your Proposition (4) is that “state coercion generally can be just and, more particularly, the state is morally permitted to impose on the citizenry to create institutions that punish (or make suffer) those who deserve it” and your Proposition (7) is that “police conduct and judicial proceedings were of sufficient quality that just punishment has not been precluded.” Yet the truth of Proposition (7) drives the truth of Proposition (4): lawless police or judicial conduct is exactly the kind of thing that can undermine the legitimacy of political institutions on the whole. Likewise, the truth of Proposition (4) drives the truth of Proposition (7): whether a state has the authority to create criminal justice institutions is, in part, a function of how those institutions behave. (To put it more informally, the answer to Proposition (4) is one of the things that determines the answer to Proposition (7), and the answer to Proposition (7) is one of the things that determines the answer to Proposition (4).)

Because Propositions (4) and (7) are linked in this way, the multiplication rule is inappropriate altogether, and the fix that you provide here doesn’t work. If we assume the truth of Proposition (4), we are saying something important about the truth of Proposition (7). If so, then there is a methodological error in treating these propositions as independent from each other. That’s not to say that the punishment theorist is off the hook, but rather that the paper’s animating methodological contribution isn’t as powerful as it might seem at first blush.

Posted by: Stephen Galoob | May 4, 2018 3:04:26 PM

Post a comment