« Liberalism for Conservatives: The Art of Separation | Main | Mootness in Sanchez-Gomez »

Monday, May 14, 2018

Reply to Rosenthal's "Response to Adam Kolber’s 'Punishment and Moral Risk'"

In Chelsea Rosenthal's interesting and carefully-crafted reply to my Punishment and Moral Risk, she is "sympathetic to the idea that we must account for moral uncertainty, and not only factual uncertainty, when determining whether we are justified in punishing" but is "not persuaded that this has the implications for retributivism that Kolber suggests." I break her discussion up into seven main points.

First, she argues that the nine retributivist propositions I examine are importantly different. Some cast doubt on retributivism, while some, she writes, cast doubt on whether particular retributivist standards are satisfied (footnotes omitted throughout):

Doubts that retributivism is the correct theory of when to punish are fairly different from doubts about whether the standards set up by retributivism are satisfied in a particular case . . . . [I]t is not clear how uncertainty about whether retributivism is the correct theory could play a comparable role in his argument—because this involves doubts about whether retributivism’s standards are the correct ones in the first place. These doubts do not suggest that punishment will often be unjustified under retributivist standards (Kolber’s claim); they are just doubts about whether to adopt those standards. Of course, if these doubts are abundant, they might, themselves, provide good reasons to reject retributivism, but this would be independent of Kolber’s argument.

My central claim is that retributivists cannot be sufficiently confident to justify the punishment of particular offenders consistent with certain values they typically hold toward the beyond-a-reasonable-doubt (BARD) standard. I don't take those values to represent a core feature of retributivism--just a common one. (Rosenthal anticipates this view in her fn. 3.) It's true that my nine key retributivist propositions are somewhat varied in their sources of error, but I don't see why that matters. In order to justly punish some particular offender, a retributivist must believe all nine are satisfied. And because confidence in all nine will always (or virtually always) be lacking or unreasonable, the paper becomes a broad indictment of retributivism.

Second, Rosenthal states "[D]oubts about the correctness of retributivism do not have to lead the retributivist to doubt whether punishment is justified. Instead, we might have someone who doubts retributivism because they have some sympathies for alternative theories of punishment, under which punishment is easier to justify." Were this to happen, I would count the paper successful. It's not as though one is a retributivist and is stuck there forever. If a retributivist begins to shift ground, he may shift to other theories and so his view that punishment is justified may stay largely the same, but he'll hold that belief for different reasons and with different implications for the nature of just punishment. The key from my perspective is that such a retributivist is not justifying punishment on traditional retributivist grounds. Indeed, such a retributivist may be following exactly the sorts of hybrid approaches that I discuss later in the paper (as Rosenthal recognizes).

Similar comments apply to Rosenthal's claim that "[u]ncertainty, for example, about whether 'suffering (or punishment) is an appropriate response to wrongdoing' raises doubts about whether retributivism is a plausible theory of punishment, rather than doubts about whether to punish a particular defendant." Agreed. I count it sufficient to raise doubts about whether an offender's punishment is justified on retributivist grounds. Maybe I needed to drive that point home more clearly in the paper, because I view it as kind of implicit. Of course, if one was a retributivist who is led to drop that view, he may be less confident that punishment is justified tout court, at least momentarily. 

Third, Rosenthal has a worry about circularity: 

Ultimately, too, using general doubts about retributivism to support Kolber’s argument has an air of circularity to it. Kolber wants to show that retributivist standards of proof cannot be met, in part because of doubts about the rightness of retributivism. This requires us to apply retributivism’s standards of proof to the question of whether retributivism is correct—in order to suggest that those standards are unlikely to be satisfied. But, I am unsure what it means to do this. We would be stipulating that retributivism’s standards are correct in order to use them to address whether retributivism is correct. The circularity is made more troubling by the suggestion that the standards would not ultimately be met—that is, that we would not be justified in adopting the retributivism whose standards we were using for the inquiry.

First, I'm not sure that the offender-specific propositions are so retributivism specific. Retributivism is a view about when punishment is morally deserved, and I think it's silent about at least some of the offender-specific propositions (e.g., whether some particular offender engaged in the acts charged, whether some particular conduct should be criminalized). Second, even if there were some overlap, it should be addressed by the same measures that I put in place to enable us to multiply confidence in the propositions: namely, each proposition assumes the truth of the prior propositions. This isn't a matter of circularity: this is a matter of deliberately assuming the truth of prior propositions to allow the math to focus on new elements of uncertainty.

Fourth, Rosenthal makes a point about forced choice which I entirely agree with. If you accept my central claim that retributivists cannot be sufficiently confident to justify the punishment of particular offenders consistent with certain values they typically hold that lead them to support the BARD standard, they could decide: (1) to drop retributivism, (2) to drop the view that punishment is justified, or (3) to drop the BARD standard or the values that I claim lead them to support it. 

Fifth, like some others, Rosenthal also wonders how closely my criticism is geared toward retributivists as opposed to consequentialists. I say that the sort of epistemic challenge I discuss potentially affects both. At least in terms of moral risk, however, I think the challenge is more serious for retributivists (who subscribe to the values underlying BARD).  For consequentialists, the standard for factual guilt is itself a matter of consequentialist calculation. So though consequentialists may be surprised by whatever comes out of the consequentialist calculation, it cannot conflict with consequentialism. It may mean that we need to punish a lot more or less than we do now. By contrast, I claim, the values underlying retributivist commitment to BARD seem to put a thumb on the scale against punishment in a way that, I claim, makes it too difficult for retributivists to be sufficiently confident to punish particular offenders.

Sixth, Rosenthal helpfully notes that the "instructional standard of proof" about factual guilt in criminal cases (BARD in the U.S.) might be different than the justificatory standard of proof as to factual guilt in criminal cases. In other words, we might really have a standard of factual guilt below BARD, but we have to use BARD for jurors to achieve appropriate results overall. I don't think this is a typical move for retributivists, but it might make sense; indeed, the values underlying BARD may demand it--given the way retributivists typically value errors of deservingness relative to underdeservingness, they might want to make it even harder for jurors to convict. The option might not be available to all retributivists, though, as some retributivists might hold standards of transparency that conflict with this approach.

Overall, though, I doubt this approach helps retributivists much to avoid my main concerns. While it might loosen retributivists up a bit in terms of the values underlying BARD, it's too insubstantial to have much effect. Suppose you think BARD is supposed to direct jurors to something like 95% confidence while our actual justificatory standard with respect to factual guilt is 90%. I don't think much changes, especially because if we're willing to punish with 90% true confidence, we've thereby introduced a lot of permissible error before getting to the tricky philosophical stuff. It will be hard to even end up above 50% confidence in the conjunction of all the propositions. (Incidentally, it's not obvious why retributivists would even have a justificatory standard with respect to factual guilt rather than just an overall justificatory standard for deservingness, but that's a matter for another day.) 

Finally, we turn to the topic of portfolios of beliefs, and I encourage people to keep an eye on Rosenthal's work in this area (see, e.g., her dissertation at her n.14). As I state and as Rosenthal recognizes, my work here is admittedly speculative and is really meant to lay groundwork for future scholarship that people might choose to pursue. So I don't think we have much to disagree with here. Rosenthal states, though:

On [Kolber's] view, combining beliefs well can reduce our risk of moral wrongdoing or help us to navigate difficult moral questions. But, at least on one natural reading, this seems to get the relationship between our beliefs and our choices backward. We may combine financial investments in ways that increase or decrease our total risk, but it does not seem that we can do this with beliefs. First, risk-reduction would be the wrong reason to hold a belief under many epistemological theories. If beliefs should aim at truth, for example, it would be a mistake to select beliefs in order reduce our risk of moral wrongdoing—and, in any case, it is not clear that we could select our own beliefs successfully.

More fundamentally, though, how risky an action is will depend upon the plausibility of moral views that condemn it; we do not adopt beliefs about those views in order to reduce (or increase) the risk. Instead, we manage our risk by adjusting our actions in light of the plausibility of different moral views.

There is, indeed, a deep debate about whether we ever should adjust our beliefs based on certain practical considerations that do not affect the truth of those beliefs. (Newcomb's paradox might be thought to raise questions of the sort.) But I wasn't seeking to take a stand on that controversial issue. In the paper, I make no claim that we ought to change our credence in particular ground level claims about morality in order to reduce our moral risk. So if I didn't rule out the reading Rosenthal considers, I simply intend to remain agnostic about it. The portfolios of beliefs discussion, however, also involves second-order claims about how we ought to think about our ground level moral beliefs. So maybe the discussion of portfolios of beliefs is meant to alter beliefs in the limited sense that some people might not have been thinking about morality using such a tool and the availability of the tool might itself alter the way people think about morality. Maybe that's the sort of belief change Rosenthal was sensing from the paper. I'm not sure. At least in Rosenthal's critique, I think she's referring to changes in ground level beliefs.

I end by expressing my appreciation for Rosenthal's deep and well-informed analysis. Among many benefits, I'm sure it will help me more clearly explicate pertinent concepts in future writing. 

Posted by Adam Kolber on May 14, 2018 at 12:07 PM | Permalink


The comments to this entry are closed.