« Prejudice Rules and Criminal Procedure Enforcement | Main | Gig Economy, California & China »

Wednesday, May 02, 2018

The Epistemic Challenge to Consequentialism

In PMR, I present the "epistemic challenge to retributivism." I argue that reasonable retributivists cannot have enough confidence in the debatable moral propositions that underlie retributivism to punish particular offenders given their commitment to the values underlying the beyond-a-reasonable-doubt (BARD) standard. 

Does the epistemic challenge apply to every purported justification of punishment? To some extent, yes. Consequentialists believe, for example, that it can be permissible to use someone merely as a means to an end if that end is good enough. How confident should they be in this proposition? I don't know. I think a reasonable consequentialist can be rather confident in it. But it might be hubristic to hold it with near certainty given that at least some of our moral intuitions seem inconsistent with it, and many reasonable people have claimed it is false. 

Nevertheless, I argue that there is an important asymmetry between retributivism and consequentialism. The values underlying BARD mean that it is generally thought far worse to punish someone who ought not be punished than to fail to punish someone who ought to be. The same values which make retributivists hesitant to punish absent solid proof of factual guilt, I suggest, should also make them hesitant to punish absent substantial confidence in moral guilt. I don't claim they need to believe the person warrants punishment beyond a reasonable doubt, but consistency would seem to require a heavy thumb on the scale against punishment. (So it's not that factual error and moral error are necessarily on a par. But they both must be taken into consideration.)

Similar comments could apply to consequentialism depending on the precise sorts of values the consequentialist holds. I'm open to the idea that the sort of epistemic challenge I raise applies to all proposed justifications and might lead some toward punishment abolition. Nevertheless, I argue that the epistemic challenge is weaker when applied to consequentialism because consequentialists are less likely to subscribe to the values typically thought to underlie BARD. They are more willing to use people merely as a means to an end when the end is sufficiently important. They worry both about the harm of making people suffer in prison as well as the harm of allowing victimizations to occur that could have been prevented. By treating the doing and allowing of consequences on more of a par, they will put less of a thumb on the scale against punishment than retributivists will. (Incidentally, the comparison I make here concerns moral risk. There is a very different sort of epistemic challenge to consequentialism concerning empirical facts that I am not discussing, even though it is an important discussion for another occasion.)

Finally, after the jump, I address a question that Asher Steinberg raised in the comments to a prior post.

An excerpt from Steinberg's thoughtful question:

What troubles me, though, about the argument is that you only argue that retributivists and consequentialists should only have these isolated doubts about certain dubious aspects of their positions, but not about their positions generally, which seem as dubious as those limited aspects. Shouldn't retributivists have moral uncertainty about the possibility that consequentialism is right generally, such that failing to punish is wrong in virtue of its causing disvalue? On the other hand, shouldn't consequentialists have moral uncertainty about consequentialism generally and therefore about whether failing to punish is particularly bad, or any worse than retributivists think it is?

For retributivists, I give nine propositions they must believe true in order to punish some particular defendant. It's true that I could have added more. So the nine propositions give us a maximum confidence. The full picture would presumably make it even harder for retributivists. (I do think I hit some of the big ones, though: free will, wrongdoing warrants punishment/suffering in return, proportionality, etc.)

Since the piece isn't focused primarily on consequentialism, the discussion there is even more abbreviated. I don't even purport to list major propositions associated with consequentialism, though again I do talk about some of the biggies: doing/allowing, means/ends. I could, for example, have had a proposition: "All else being equal, some good consequence is better than a less good consequence," though that would attain nearly universal support.

Importantly, however, the project is not meant to be a full-fledged bottom-up analysis of the appropriate credences of each purported justification. I take as a starting point a reasonable retributivist and a reasonable consequentialist. So I grant, at least for the sake of argument, rather high confidence in the major beliefs of each. Still, I argue, at least in the case of retributivists, those relatively high confidences compound in such a way as to create rather low levels of confidence relative to what they would likely demand (if they subscribe to the values that seem to underlie BARD).

Lastly, retributivists who purport to justify punishing some particular offender by some particular amount needn't worry whether consequentialism would also justify that same punishment. It would just be more reassuring, perhaps. I do think retributivists need to worry about failing to punish when there would be better consequences if they did punish, but I don't add such propositions in for the two reasons already given: (1) I'm discussing the maximum confidence in retributivist punishment and these other propositions certainly could have been added to make the confidence lower, and (2) I'm willing to grant that reasonable retributivists will have relatively high confidence in the negation of key consequentialist premises and vice versa. But I'm certainly open to other ways of conducting these analyses, each of which provides somewhat different but still quite valuable information. Also, I have a discussion in the paper about epistemic hybrids in which consequentialists do think about retributivism and vice versa. That portion of the paper may also be relevant to your comment.

Thanks, Asher! (Some of the propositions in the post above get challenged by the respondents, so I'll be returning to them again over the next couple of weeks.)

Posted by Adam Kolber on May 2, 2018 at 12:27 PM | Permalink


Regarding your second paragraph, I think you make a good clarification. I meant that I grant high levels of confidence in particular retributivist propositions, but you're right that I claim that, even after doing so, reasonable retributivists will not have sufficient confidence in the ultimate conjunction.

Re: "Now, maybe this exercise shows that it makes no sense to be a retributivist (which is what you claim you aren't arguing). . ."--> I do, in fact, take the paper to present a challenge to retributivism. (So not that retributivism literally "makes no sense," but I do take the paper to be a serious challenge to traditional forms of retributivism.)

Re: "I don't think it shows that *actual retributivists* should be paralyzed by moral doubt from punishing anyone because actual retributivists aren't 80% sure of this and 80% sure of that and so on, such that they're only 80% confident to the ninth power of the whole theory, or less than 10%." I'm not sure what you mean exactly. Maybe retributivists don't spend time thinking about their levels of confidence, but they need to do so. They can't be confident in retributivism if they're not confident in the propositions required to retributively punish particular offenders.

Posted by: Adam Kolber | May 3, 2018 7:21:25 PM

I appreciate the appreciative comments on my comment. I haven't read the responses, but I skimmed through Chad Flanders's, and I think there's a lot to what he says about Solum's critique. I've always been struck by Rawls's invocation, at the end of his preface to Political Liberalism, of Kant's remark that if justice perishes, then it is no longer worthwhile for men to live upon the earth. Helpfully to your characterizations of retributivism, Kant actually said that in the context of denouncing consequentialist justifications for punishing the undeserving; if we unjustly punished, he grandiosely and apodictically claims, human life would lose its purpose. But the extraordinary value Kant places on justice is a two-sided coin. In the prior sentence, he writes, "The penal law is a categorical imperative, and woe betide anyone who winds his way through the labyrinth of the theory of happiness in search of some possible advantage to be gained by releasing the criminal from his punishment or from any part of it, *or* [emphasis mine] who acts in the spirit of the . . . saying, 'It is better that one man should die than the whole people should go to ruin.'" And then he goes on to posit the case of a condemned criminal whose life is spared so that doctors can perform potentially fatal medical experiments on him, to which his objection seems to be the injustice of possibly sparing his life ("justice ceases to be justice if it can be bought as a price") rather than the cruelty of his use. Failing to punish the deserving, just as much as punishing the undeserving, is something so to be dreaded for Kant that he'd rather give up on humanity altogether. Kant isn't, of course, a contemporary retributivist and today's retributivists are less apt to explicitly fret about the injustice of failing to execute or incarcerate the guilty, but I wonder if that's more a cultural difference than a difference in contemporary retributivist theory from Kant. So I don't know that it's true that the consequentialist costs of failing to punish have greater weight than the retributivist costs.

On your response to my comment, you say you're assuming people with high confidence in retributivism and consequentialism, and you purport to not be engaged in a bottom-up analysis of the theories themselves. But it strikes me that what you've really done is assumed someone with high independent levels of confidence in propositions that make up retributivism, which multiply to a low level of confidence in the theory as a whole. If you really want to talk about someone with high confidence in retributivism, don't you have to stipulate that that's just what they've got, and then reverse-engineer from that really high confidence in the claims that comprise the theory? As the argument stands, it looks to me like you're only playing at charitably assuming high confidence in retributivism; you allow the retributivist high, perhaps unreasonably high, confidence in free will and so on, but not so unreasonably high that he can maintain confidence in the theory as a whole once you're done multiplying out all the confidence levels. Now, maybe this exercise shows that it makes no sense to be a retributivist (which is what you claim you aren't arguing), but I don't think it shows that *actual retributivists* should be paralyzed by moral doubt from punishing anyone because actual retributivists aren't 80% sure of this and 80% sure of that and so on, such that they're only 80% confident to the ninth power of the whole theory, or less than 10%.

Posted by: Asher Steinberg | May 3, 2018 6:57:02 PM

Sam and Stephen: Thanks!

Peter: These are good points and Emad Atiq's response goes in a similar direction. A few things to say now:
* You're absolutely right that nothing about commitment to BARD *entails* one's view about how high a justificatory standard of proof must be. I frame the argument in terms of commitment to certain values and that raises the question of what exactly those values are.

* The piece tries to speak in terms of the rationality/consistency of retributivists so some of the consequentialist reasoning you give may not, as an empirical matter, represent the sorts of reasons retributivists are likely to give.

* More importantly, I do address rather head-on some of the concerns you raise. To be sure, there are reasons to have a high standard of factual guilt that do not apply automatically to one's overall justificatory standard. But I do press on the consistency of having radically different standards. For example, in some jurisdictions, the BARD standard would apply to the question of whether a person acted as an automaton who made no voluntary action. Why apply such a high standard to that question but settle for something dramatically less as to the question of whether we ever have free will?

* You write: "One could have the deontic view that it's simply necessary to establish the factual elements of the crime with the given degree of certainty." True, I'm just pressing on the reasons a retributivist might give in defense of this view.

* You write: "To put the point differently the values underlying BARD might well be that we should be especially careful not to punish people who have behaved well and properly (either because consequentially it is super bad or just as a matter of moral [deserts])." Again, fair enough re: consequentialism. But moral desert is negated for retributivists no matter whether its due to lack of factual guilt or lack of moral guilt. It's hardly obvious why retributivists are so demanding with respect to factual guilt relative to moral guilt. So, I'm sure there are implicit assumptions in the blog post, but I do try to address these concerns, at least in an abbreviated form, in the paper itself. But you could also read the paper as calling out for a retributivist explanation of its apparent differential treatment of factual and moral guilt. Thanks for your very helpful contribution!

Posted by: Adam Kolber | May 3, 2018 3:45:30 AM

To put the point differently the values underlying BARD might well be that we should be especially careful not to punish people who have behaved well and properly (either because consequentially it is super bad or just as a matter of moral desserts). As such it doesn't follow that we need to be particularly careful not to punish someone who has behaved badly (violated announced laws etc..) if they don't deserve it.

It could well be that BARD applies to how confident we must be they did the deed and then we just go by preponderance of the 'evidence' regarding whether or not someone we are BARD convinced is guilty deserves punishment.

Posted by: Peter Gerdes | May 3, 2018 12:02:36 AM

It seems to me that you are making some rather substantial assumptions about the nature of our commitment to BARD. In particular, you seem to assume we assent to some meta-moral principle of the form: unless we are convinced to more than X degree that someone deserves punishment for a crime we shouldn't impose it.

There is nothing at all inconsistent about believing that the BARD standard is appropriate for the individualized evidence of having committed a crime but not required regarding the moral appropriateness of the general rule demanding punishment. The most obvious way out is to be a kind of consequentialist who adopts the BARD rule regarding courts and trials because of a belief that it maximizes utility to give courts this rule.

However, one need not be a consequentialist to take a similar position. One could have the deontic view that it's simply necessary to establish the factual elements of the crime with the given degree of certainty.

Also, it seems plausible that very few beliefs in our moral portfolio actually deem it morally wrong to punish most offenders.

Posted by: Peter Gerdes | May 2, 2018 11:54:57 PM

Interesting! Might usefully be contrasted with Lenman's "Consequentialism and Cluelessness," https://doi.org/10.1111/j.1088-4963.2000.00342.x .

Posted by: Stephen Sachs | May 2, 2018 9:45:47 PM

It sounds like James's understanding of an intellectual pursuit worth writing about reduces to only ever attempting to give answers to questions other people have already asked, or are likely to ask given some set of priors within the type of intellectual-but-not-too-intellectual sort we typically find among law school faculty. Of course, I'm sure the 439th article on whether Chevron has 1 step or 2 steps or 1.234 steps is "useful" in the kind of way we can get on board with.

Posted by: Sam | May 2, 2018 9:22:46 PM

James: I don't understand. Are you saying that there aren't people who confidently believe in retributivism?

Posted by: Adam Kolber | May 2, 2018 7:33:07 PM

Your argument against the high confidence retributivists reminds me of Davidison's argument against "the general skeptic". The problem with both these arguments is that it is difficult to find anyone who actually subscribes to such beliefs. Yes, yes every so often a person shows up in philosophical circles claiming they are skeptical of everything including their skepticism and usually people laugh at them and go back to bed. Where are these people who would only find consequentialism "more reassuring"? Outside of purely academic debates?

I basically see you launching an intellectual attack on a bogeyman. Which would be quite useful if the world was populated with bogeymen, but it is not.

Posted by: James | May 2, 2018 6:30:08 PM

The comments to this entry are closed.