Kearns and Beebee Reply to Smilansky

In this month's issue of Analysis, there appear two articles replying to Saul Smilansky's piece from last October, "Determinism and Prepunishment: The Radical Nature of Compatibilism", which we discussed on the Garden here. One is by Stephen Kearns and the other by Helen Beebee (subscription required to access). Happily, the issue also contains replies by Smilansky to each of his critics.

(Also in the issue is Joe Campbell's rejoinder to Brueckner's reply to Campbell's article about the Consequence Argument, which we have also discussed at the Garden here. In fact, both Campbell and Smilansky cite the Garden in footnotes!)

The articles are all quick reads, so I recommend them to everyone. However, just for the sake of discussion, I provide summaries and my own opinions below the fold.

Kearns has two main problems with Smilansky's argument. First, he points out that many compatibilists maintain that even determined wrongdoers nevertheless have the ability to refrain, and thus that we should still allow them the opportunity to exercise that ability. Second, he points out that there are imaginable circumstances in which the prepunishment is actually what causes the punished person to commit the crime at some later point in time. In these circumstances, Kearns maintains, prepunishment is surely morally abhorrent, and the compatibilist can agree.

In his reply, Smilansky responds to Kearns' first point by asking what we could possibly be waiting for when we say that we are allowing future wrongdoers the opportunity to exercise their ability to refrain from committing the crime. They *will* do it, and thus it would seem pointless to wait around. In response to the second point, Smilansky says that the circumstances imagined by Kearns are irrelevant to his main argument, since the sorts of case Smilansky had in mind did not involve the prepunishment causing the future wrongdoing.

It seems to me that Smilansky's responses here are decisive. Moreover, an additional way to respond to Kearns' first point, it seems to me, is simply to say that he is conflating ability with opportunity (even when 'ability' is construed in a compatibilist-friendly way). Kearns says: "According to many compatibilists, even if a person is determined to do something, she is still able to decide not to do this thing. Of course, she won't decide not to do it, but she can. Thus before she commits a crime she is still able not to do so. Even if she firmly intends to commit the crime, she is able to change her mind. Therefore, she still has the opportunity to remain innocent" (p. 251). But it seems to me that the relevant sense of 'opportunity' at play in the claim that we ought to allow even future wrongdoers the opportunity to refrain is a sense that is not entailed by any compatibilist-friendly construal of 'ability'. So Kearns' move here from the agent's ability to refrain to her "opportunity to remain innocent" seems either fallacious or at least involving an unhelpful sense of 'opportunity'.

Beebee also has two objections. First, she points out that what seems to be driving Smilansky's argument isn't determinism but predictability instead. What seems to bother Smilansky is the mere fact that there is a determinate fact of the matter, ahead of time, about whether someone will commit a crime. But even some libertarians are committed to this, so there is no special problem for compatibilism here. Her second objection involves an appeal to the way we punish people for conspiracy to commit murder -- it does amount to prepunishment in certain circumstances, she argues, and so it is unclear that prepunishment is always so morally abhorrent.

Smilansky replies to her second objection by maintaining that even when we punish people for conspiracy to commit murder, this does not count as prepunishment -- we aren't punishing them for the act that they were planning to (but didn't) commit. Rather, we are punishing them for something they did do in the past, namely conspire (or something along those lines). This seems a persuasive response to me.

I am less persuaded, however, by Smilansky's response to Beebee's first objection. He says that that mere predictability is not sufficient for prepunishment. Rather, "the reason why we may punish at t0 is not predictability but that the crime, in a sense, is already there" (p. 261). And the crime is "already there", according to Smilansky, because the intention to commit the crime is already there, and it is determined that that very intention will cause the criminal action at some point in the future. As he puts it on the next page (p. 262), in these circumstances, the crime "deterministically pre-exists". On a libertarian view of things, on the other hand, "the crime obviously cannot already be there, at t0, because it is going to be caused, through the exercise of LFW, only later" (p. 262).

I may be missing something in Smilansky's response here, but it seems to me that he is drawing a distinction without any metaphysical basis. Suppose just for the sake of simplicity that eternalism is true: that is, suppose that past and future objects exist, as well as present objects. (The past and future objects do not exist now, of course, but the eternalist does not think that such a property is required for existence.) Now consider two scenarios: one in which determinism and compatibilism are both true, and in which someone has at t0 an intention (which will be causally efficacious) to commit a crime at t1, and the other in which libertarianism is true and in which someone has at t0 an intention (which will be causally efficacious, though not deterministically so) to commit a crime at t1. If prepunishment is permissible in one of these scenarios, I would think it should be permissible in the other, as well. Smilansky seems to think that the crime in the first scenario is somehow more robustly "already there" at t0 than it is at t0 in the second scenario, but I can't see how to understand this claim. It's certainly true that it is determined in the first scenario and not in the second, but how does this make it more "already there"?

In any case, I'd be interested to hear what others think about this issue. It's interesting, and I'm glad for all four of the Analysis articles, which have given me the opportunity to think more about it.

UPDATE: I went back and read some of the comments in the thread where we discussed Smilansky's original paper, and I think Smilansky may simply reply to my last worry by denying eternalism. He seems to suggest this in his comment on October 3, 2007 at 10:15am. It would be interesting if the worry about prepunishment depended upon some particular view about the ontology of time.

Comments

[Mark Smeltzer]

Neal,

Thanks for the heads up. I rather enjoyed the discussions that emerged in the previous thread on this subject.

In the prior thread, I argued that ALL of Saul's cases involved some sort of real liability in the past before the crime was committed, and that compatibilists could explain a non-abhorrent right to punish by appealing to those guilt conferring facts. The guilt would certainly be limited to the scope of the facts, but this liability promises to be real, non-abhorrent punishment-entailing guilt nonetheless.

Saul replied by saying there are other cases that aren't affected by the "over moralization" that I latched onto in the cases he offered, but he did not specify any of those cases to backup his point.

One of the movies I saw recently (though I won't say which one and spoil it for anyone) contained a case that may illustrate some of the points that Saul was making later in that thread:

1. A virtuous man's beloved fiance is brutally murdered.
2. The sheer violence and cruelty of this act erodes the man's foundations for upholding virtue.
3. This loss of "faith" in virtue triggers a violent restructuring of his values.
4. His new, warped values lead him to become a murder himself.

It seems unquestionable that it would be morally abhorrent for us to punish this man before his fiance is killed, and completely warranted for us to punish him after he commits to committing his first murder. The question I think Saul presses is how (and where) can the compatibilist draw the line between these two extremes.

Does this seem a fair way to sum up Saul's challenge to compatibilists? I hope I'm not over simplifying things.

[Clarissa Millarker]

Without having access to the articles and relying on Neal's summary, it seems to me that Kearns is on to something. If retributive punishment is justified at all, it seems that it is justified as a response to behavior. However, if carried out prior to the naughty behavior, it seems inappropriate to consider it a "response" at all... even in the deterministic case in which an intention (to do wrong at t1) is guaranteed to be efficacious in the future when it arises at t0. In the game of Whack-a-Mole that is retributive punishment, bonking a mole who has not yet emerged from his mole-hole gets you no points.

Very limited "punishment" of a utilitarian character (to prevent wrongdoing or protect people from wrongdoing) seems okay, however, and it seems that Kearns has at least a signpost to this insight when he points out that prepunishment could actually be the cause of blameworthy behavior in some cases. Any punishment that causes the very bad behavior it is meant to stamp out is sure to be rejected by utilitarian reasoning. If prepunishment can prevent wrongdoing, however, the utilitarian would apparently have to conclude that we should prepunish in that instance.

This, however, leads to a fresh puzzle: if prepunishment prevents the wrongdoing, then for what is the agent being prepunished? The utilitarian might simply shrug and respond that the consequences of prepunishing (or, performing the act that would have been prepunishing if the wrong-doing had been allowed to occur) are preferable to the consequences of refraining from prepunishment.

However, this line of reasoning toward prepunishment that prevents wrongdoing seems slightly suspicious to me, because it seems that there is no possible world in which all of the following are true:
1) W entails (if P, P is a response to W)
2) ~W entails ~(if P, P is a response to W)
3) P entails ~W
4) W entails ~P

In a deterministic world where P prevents W (and the non-occurrence of W is not overdetermined, naturally... ~W's being overdetermined would defeat W's serving as a justification for P, anyway), surely 2 and 3 are true. 2 is simply the denial that one could be "prepunished" for something one will never do (which seems true), and 1 is what I take to be a plausible translation of "P counts or would count as prepunishment for W."

This is not strictly a problem for the utilitarian, since the utilitarian should simply choose to act so as to bring about the best possible state of affairs (and so, choose P without regard to any counter-factual weirdness surrounding W.) However, for any compatibilist who demands that the agent deserve P, this looks like a serious problem... Unless, of course, we allow ourselves imaginary worlds analogous to the imaginary numbers to which mathematicians help themselves. It's a shame David Lewis left us so soon.

Obviously, I've shot from the hip, here, but I hope someone can come along and make some sense of this.

[Matt King]

Since Neal brought this up again, I thought I might post my response to Smilansky, since I think it's simpler than either Kearns' or Beebee's. I'm curious to see what the gardeners make of it.

Smilansky's making two main claims.
(1) Compatibilists can offer no principled objection to prepunishment; while,
(2) Libertarians can offer an intuitively plausible principle: respect for persons forbids us from prepunishment so as to allow the individual to change his mind.

Thus, Smilansky's charge is both that compatibilists can't offer any good principle, *and* that libertarians have a very good one to offer.

But I think it clear that compatibilists do have a principle they can offer; indeed, it's one that anyone can offer, no matter one's affiliation. And, I think, it's just as plausible as Smilansky's principle.

I call it: Just Punishment Requires Responsibility (JPRR). JPRR states that an individual S can only be justly punished for x, if S is responsible for x.

JPRR is a highly intuitive principle; I would argue at least as intuitive as Smilansky's own. And compatibilists can clearly offer JPRR even in the cases Smilansky examines. Just because it is determined that S will do x at t1, even if S now (at t0) has most of the mental states which would be relevant to explaining his responsibility for x, it doesn't follow that he *is* yet responsible for x. And if he isn't responsible for x, then he cannot be justly punished for it.

At most the compatibilist is committed to the claim that S *will* be responsible for x at t1, and that perhaps we could know this with 100% certainty (or beyond a reasonable doubt, as Smilansky emphasizes). Smilansky wold have to show that because S will satisfy compatiblist conditions on responsibility at t1, he satisfies them now. But I know of *no* such compatibilist who holds such a view, and compatibilism as such certainly isn't committed to such a view.

Smilansky, indeed, anyone, might of course object to JPRR. It might not be true. But it seems to me a perfectly plausible candidate. Indeed, part of its plausibility seems stem from its *not* taking a stand in the dialectic.

[Neil]

Matt, you can't just stipulate the only after the event the agent will be responsible for it; you need to say in virtue of what the agent will be responsible for it. A natural possibility is to say that we need to wait to give the agent the opportunity to do otherwise. But well-known difficulties lie in that direction.

Neal, Smilansky may reject eternalism, and for all I know he may be right to do so. But predictability - beyond reasonable doubt - does not depend on eternalism, and all libertarians are committed to the view that there are some morally significant actions performed by actions that at least indirectly free and for which the agent is responsible. So Beebee seems right in thinking that prepunishment is a problem for everyone.

[Cihan Baran]

Hello,

It's best to remember the most troubling case for pre-punishment, which might also be considered Smilansky's strongest case in my opinion. If I remember correctly, it went as follows:

Case X: An agent A is going to commit a crime C at time t_c. We are unable to punish A after he commits the crime at t_c or prevent him from committing the crime at t_c. Moreover, our punishing A before t_c will have no bearing on his deliberations and/or motivations for C. A will go onto commit C whether we prepunish him or not.

JPRR may be highly intuitive but in Case X, doesn't it actually amount to letting go of the agent?

The case I have in mind is troubling in two respects - it's like a moral dilemma. If we don't pre-punish the agent, we sort of allow him to do what he does unpunished. If we pre-punish the agent, we give him a moral cheque to commit the crime - we sort of legitimize his criminality.

Moreover, I thought that the whole appeal of Smilansky's prepunishment argument was the idea that if compatibilism is true, I am already responsible for everything that's in my future - not that I will gain responsibility when I commit the act but I already am responsible, since everything is determined.

Lastly, I don't see how prepunishment is troubling for the libertarian. Yes, there may be predictable acts in a libertarian view but the moral responsibility granting ones will be acts that are unpredictable - such as Kane's SFAs.

[Matt King]

Neil,

I'm not *stipulating* that the agent is responsible only after event. But if my compatibilist conditions on responsibility aren't satisfied prior to the action, then I'm under no pressure to grant that the agent is responsible prior to the action. If Smilansky's argument is going to work, it has to work independently of the details of libertarianism. It has to force us to accept libertarianism *because* of pressure brought on by his examples. But if I'm right that JPRR applies to his cases, then I've provided a plausible principle, one that's pretheoretically appealing, contrary to his claims.

[Neil]

Matt, you've said that the agent is not responsible prior to the event but is responsible afterwards. But you haven't said why. That looks like stipulaton to me. In fact, I accept JPRR. I guess Smilansky does too. The question is, when is the principle satisfied? You say after the event; Smilansky says that given compatibilism JPRR is satisfied now. Invoking JPRR doesn't settle this, without saying what waiting accomplishes (ie, answering my 'in virtue of' question).

[Matt King]

Cihan,

I don't think JPRR amounts to "letting" the agent go in Case X. I'm comfortable granting that there may be cases in which we can't punish people, even though we may like to. And we will be warranted in condemning him, so long as he's responsible.

Also, I don't think Smilansky has established that we're responsible for everything we're determined to do. That verdict is tied into his being right about his cases. Obviously, I think he isn't right about his cases.

[Matt King]

Hi Neil,

It's in virtue of your favorite compatibilist conditions on responsibility. So far as I can figure, no extant account is committed to attributing responsibility before the event even in the cases Smilansky describes.

Smilansky's free, of course, to claim that in his cases, there's no good reason to deny that the agent is responsible. But if one's established compatibilist account rules that the agent isn't yet responsible, I see nothing in the cases themselves that ought to push me to give up my compatibilism.

I see nowhere in Smilansky's argument where he shows that a compatibilist account would necessarily be met at all times if determinism were true. That's a bizarre claim for anyone to make, I should think.

Smilansky believes that if we know that S intends to do x, and will certainly continue to intend to x, and that he will certainly in fact x, that we have all we need to ascribe responsibility. And he may be right that, epistemically, we do have all we need. But this fact won’t amount to S actually being responsible. All we’ll know is the facts that determine she will be responsible; not that she is.
And JPRR suggests that whatever facts determine responsibility, some of them will be facts having to do with actually committing the offense. Smilansky can, of course, reject this claim, but I
think he bears the burden of showing why this radical position on responsibility, and
prepunishment in general, is more justified than the ordinary notions supported and expressed by JPRR.

In the interest of full disclosure (and shameless self-promotion), I've excerpted this last bit (with minor alterations) from my short paper on the topic, posted on my website: wam.umd.edu/~mattking.

[Neil]

My favorite compatibilist condition on MR is silent on tense. Suppose it is in virtue of moderate reasons-responsiveness (that *is* my favorite). Well, since we can know that the agent will be MRR when they act, it is directly question-begging to cite this condition against Smilansky, without adding something about what waiting adds.

[Alan]

Calvin in his Institutes first proclaimed the doctrine of double predestination--which of course includes damnation of specified individuals before (in some intelligible sense of "before") the temporal fact. This is because the future is not only fixed by the predestinator, but planned as fixed. (Those are of course logically distinct points.) Presumably Smilansky's prepunishment supposes a Calvinistic God's perfect state of knowledge of future transgressions, and given that chaotic epistemic indeterminism is entailed by finite knowledge of the future even in the case of metaphysical determinism, there are empirical limits to claims that such transgressions will in fact occur. But if infinite knowledge of such chaotic actions of the future is supposed (as the Calvinistic God might have), then (arguably) one would possess knowledge of the power that would (counterfactually at least) provide avoidance of that transgression. If one could locate that power (temporally indexed as prior to the transgression) in the future transgressing individual, then prepunishment would make some ultimate sense (as with Kane's SFAs); if not, then such ultimate responsibility moves back in the Strawsonian sense infinitely (and/or nonsensibly) or attaches to some being like the Calvinistic God who then absorbs moral accountability (and who, given the circumstances, looks to be pretty arbitrary at least to finite minds). So, Calvin's God aside, prepunishment only makes sense for libertarians like Kane who can locate ultimacy within an individual, and for such individuals we can somehow know will perpetrate those known transgressions. Empirically then, epistemic finitists--as are most compatibilists--are off the hook for prepunishment.

[Mark Smeltzer]

Alan,

I am pretty sure that Augustine got there first with the Doctrine of Original Sin.

Regardless, it is well known that the entirety of the Reformed Christian tradition holds to the concept of preguilt and predestination based upon whether an agent is judged in its own naturally guilty state, or whether the agent's guilt has been pardoned according to Christ's imputed righteousness -- i.e. on the basis of whether the agent's guilt has been forgiven. Thus, even stillborn babes could plausibly stand before the judgment seat.

As you mentioned, the God of the Reformed Christian tradition is wholly unlike you or I: perfect judgement derived from perfect eternal knowledge derived from the perfect realization of an eternal plan. However, I strongly doubt that Smilansky has anything like the God of Reformed Christianity in mind since he discusses (what he believes t be) the plausibility of future devices that will allow us humans to foreknow the acts of other agents.

In the prior thread we discussed a strong reason to believe that such devices are metaphysically impossible in the actual world (even if the actual world is deterministic) given the world's chaotic state. I still believe there is merit to the "plausibility" objection to Saul's cases, but I don't see what's to stop him from positing the same cases and stipulating that "the events in these cases occur in deterministic, yet non-chaotic, worlds."

In other words, I believe that the foreknowlege barrier by itself does not impose on the logical possibility of prepushishment for the Compatibilist. However, it sure would stand in the way of prepunishment ever being a morally abhorent threat...

[Neil]

Right Mark, but we don't need perfection - not if the legal standard of guilt picks out a threshold for the appropriateness of punishment. It is, as Smilansky points out, quite reasonable (whatever the causal structure of the universe turns out to be) to think that we can now or soon will be able to predict what some agents will do in some circumstances beyond reasonable doubt.

[Alan]

Mark--

Augustine got there first by damning everyone by original sin--but not as Calvin said by putting in the fix beforehand: Augustine for better or worse relied on some sort of FW for temporal redemption; Calvin said in an eternalist sense: fuggetaboutit.

Mark and Neil--

But my point is what would count as empirically reasonable doubt? Logical possibility would not do--or else OJ could have claimed he was abducted by aliens while Nicole was nearly beheaded. Given the empirical constaints of chaos theory, what then could reasonably acquit a juror of his or her reasonable doubt about what a person could have done? Look back on my comments about God's perfect knowledge--knowledge of what one would do in even chaotic deterministic circumstances might entail power/ability-involved claims of what one might have done otherwise, but in that case we encounter uncertainties about whether we are blaming the actual individual under scrutiny (these are famous problems of counterfactuality under the assumption of determinism), and these uncertainties seem to extend to non-chaotic deterministic worlds, and if the individual chooses indeterministically, then there are uncertainties of a distinctly metaphysical character. In the former case we might have real philosophical difficulties about whether we are justly accusing the very same deterministic individual; in the latter case we might justly accuse the indeterministic chooser--if we can then solve the accompanying problems of ultimacy. (L's black box.)

Prepunishment is not an easy and given conceptual problem. Smilansky is to be credited for raising the issue--but its resolution is no slam-dunk by any means.

[Neil]

Any doubts you have about reasonable doubt, Alan, seem to apply to the legal standard as well (at least I don't see why not). Do you think that we just empty the jails? (cheap rhetoric? No: *I* think we should empty the jails. But that's another story).

[Mark Schroeder]

Neil - Sorry to come late to the party, but I doubt that your favorite compatibilist condition on MR is silent on tense. Consider this view: that you are MR for an action up until exactly three months, seven days, and twelve minutes before you do it under your moderate reasons-responsive condition, and thereafter you are _not_ MR for it.

I doubt that you would accept this as an acceptable precisification of your favorite compatibilist condition. If you wouldn't accept it, then your view is not silent on tense. What is more likely the case, is that your view is _non-explicit_ about tense, rather than that it is silent about it.

[Neil]

I doubt we can deduce that my favorite view has implicit commitments on tense from the claim that there are positions regarding tense that are crazy. That aside, anyone who wants to claim that compatibilism can resist prepunishment must do a lot of work, even if that work consists in making the implicit explicit. Its not sufficient to cite a putative necessary condition, when it is open to prepunishers to claim that agents satisfy that necessary condition in a tenseless manner.

[Mark Schroeder]

Neil - I'm sorry; I thought the argument was about whether there are plausible principles that real-life compatibilists can appeal to in order to resist prepunishment, not whether there are or are not possible compatibilist theories which can't.

Matt pointed out that he knows of no existing compatibilist account according to which the bare fact that someone will be MR for doing something entails that she is already MR for doing it. It follows from Matt's observation, if it is true, that unless there is some important account Matt has failed to think of, no existing compatibilist is on the hook. So I don't see what the problem is.

Look at it this way: typical analyses of what it is to be square don't mention tense. They don't say, "to be square at time t is to have four equal sides at right angles at time t". They just say, "to be square is to have four equal sides at right angles". Moreover, it is a plausible principle that you shouldn't put a square peg in a round hole. Now, Saul Smilansky comes along and points out that it is possible for there to be square pegs that we can know in advance will be manufactured tomorrow into round pegs, and challenges our theory: "your reductive theory about the nature of squareness can't explain why there is something wrong (today) with putting those pegs into round holes! But clearly, he advises us, there is something wrong (today) with putting those pegs in round holes. So we should prefer a theory with different resources for explaining that. But so what? Just because the pegs will be round tomorrow doesn't mean they are round today. Today they are square, so the principle applies: don't put them in round holes. Tomorrow it won't apply, and we can know that in advance, but that doesn't change what the rule lets us do today.

So here's the moral: independent plausible principle meets reductive theory of squareness, and makes concrete - and intuitive - predictions. The reductive theory of squareness didn't make explicit claims about times, but there were charitable ways of understanding it and uncharitable ways of understanding it. It's uncharitable to understand it as holding that something is square at time t just in case there exists a time t* such that it has four equal sides at right angles at t*.

You say: but the pre-punisher can just give an intepretation to the theory of MR on which it is atemporal! That's like interpeting out theory of squareness in this uncharitable (and obviously false) way - a way which, as Matt notes, is accepted by no existing compatibilist theory he can think of. Why should Matt be any more worried than my theory of squareness?

[Neil]

One of us has lost their grip on the dialectic: there is a non-zero possibility that it is me. As I see things, Saul says that given compatibilism and the knowledge (or highly probable belief) that an agent will X at t, there is no point in waiting until t to punish. You say there are plausible principles the compatibilist can appeal to resist this conclusion. Well maybe, but I want to know what they are (that's what the debate is about, as I see it). It can't be *responsibility is necessary for punishment*; Saul claims that given compatibilism (etc) that principle is satisfied. In other words, my claim wasn't that there are possible tenseless compatibilisms, but that Saul has made case for the view that compatibilism is essentially tenseless.

[Mark Schroeder]

I don't see how Saul has made the case that compatibilism is essentially tenseless. Nor do I see how he could. Here is a compatibilist view that is tensed: X is responsible for A just in case (1) X brought about A and (2) X did so in the black-boxedly-responsibility-involving way. (My theory then goes on to spell out the contents of the black box.) Prior to A, my condition (1) is not satisfied, so my account predicts that prior to A, X is not responsible for A. Yet my account is clearly compatibilist, as long as the contents of my black box are compatibilist. Saul can't have argued this view out of existence, so he can't have made the case that compatibilism is essentially tenseless.

Any compatibilist theory which shares the features of this view - that you can't be responsible for things that haven't happened yet - will share with this view Matt's response to Saul - that just punishment requires responsibility. It appeals to a genuinely plausible independent principle (JPRR), and to a plausible principle about responsibility which is denied by no compatibilists Matt can think of, and it answers the challenge.

Now, you might think that I'm being dense, because the more interesting question is _why_ you can't be responsible for something that hasn't happened yet, and moreover, _why_ JPRR would be true, if responsibility is backward-looking rather than forward-looking, in this particular way. I agree - those are great questions. But it still looks like Matt is giving us a pretty obvious place to look. If these principles really are independently plausible, then they can explain why prepunishment is unjust, even if we will also want to explain them, in turn. They still give us a simple answer to Saul that many theorists should be able to agree on. And the simplicity and ecumenicality of the answer was the virtue Matt advertised for it. No?

[Ryan Lake]

I think it's helpful to recall that the prepunishment debate started in 1992 with Christopher New actually asserting that there is no reason to wait until the crime has occurred to punish - especially if the agent has already formed the intention, we know in advance that he WILL commit the crime, and there will be no opportunity to punish after the crime. In such a case there seems to be at least some prima facie appeal to prepunishing, because unless we do it now, before the crime has occurred, there will be no opportunity to make the agent pay for his crime.

In arguing that respect for individuals requires giving them the opportunity to remain innocent of the crime, Smilansky is attempting to explain exactly why New is wrong - why it isn't ok to prepunish even in such cases. Only later did Smilansky note that this reply is apparently not available to the compatibilist, leaving the compatibilist with the challenge of offering a different explanation of what exactly is wrong with prepunishment.

So the challenge is respond not to Smilansky, but to someone like New, who is arguing that prepunishment really should be a viable option. Without more explanation, Matt's appeal to JPRR just seems like stamping one's feet and asserting that prepunishment isn't ok. A prepunishment proponent like New isn't going to accept the bare assertion that temporality is relevant to responsibility without some explanation why. Of course maybe that's not a bad thing in this case - maybe prepunishment is so bizarre and so obviously wrong that foot stamping is the appropriate response.

[Alan]

I just finished writing about 1500 words to post here, but realized I was just prepunishing Gardeners as I was punishing myself. So you're off the hook.

A short summary instead. The puzzle of why shouldn't compatibilists prepunish really seems predicated on discussing desert-entailing, ultimacy-grounded, retributive-type responsibility. If we replace that with deterrence-rehabilitation non-ultimacy, non-retributive-type responsibility that is typically associated with classic forms of compatibilism, then given the condition of assurance about a kind of transgression that will occur as stipulated above (I still am uncomfortable with that assumption in many ways, but I realized that this has to do with my excessive worry about knowing specific future events as opposed to types of events), then why couldn't a compatibilist simply agree that they *should* prepunish? That would be because what is termed "prepunishment" here is otherwise understood by the compatibilist as *prudent prevention* of an anticipated kind of transgression? Understood that way, the accusation of having to engage in "prepunishment" is robbed of all real force against the compatibilist, because presumably anyone would want to act so as to prevent future evil they were relatively sure would take place otherwise.

This has been a great discussion and has nagged me immensely--but now I'll stop punishing everyone including myself and shut up!

[Mark Smeltzer]

Alan,

This is a side bar: I've always thought that Augustine's debates with Pelagius confirmed his stance on FW, and his commitment to Theological Determinism. Regardless, another interesting point in Reformed history is Leibniz's view of determinism and responsibility -- he was certainly on the same page as Calvin. And then there's more modern examples like Jonathan Edwards.

All of these examples go to the point that the Reformed Christian tradition embraces the idea of preguilt. However, I'm still not sure if this is relevant to the concept of prepunishment.

[Matt King]

Ryan,

I don't think appealing to JPRR is stamping one's foot. I think it *is* an explanation for why prepunishment is wrong, even in a New-type case. It's because the agent involved isn't yet responsible for the crime, even though he will be (and we know this).

Alan's right that JPRR, as I understand it, appeals to a desert-entailing notion of responsibility and prepunishment as a form of retributive (desert-based) punishment. A discussion of prepunishment would surely be broader than my focus with JPRR.

[Matt King]

Sorry Ryan, in reading over my comment I realized I hadn't fully addressed your worry. I'm not sure whether temporality is, strictly speaking, relevant to responsibility. I say more about this in the paper on my website.

But I will add this, New seems to think that our unease at not punishing someone when we can (giving up the chance to punish, as it were) will outweigh any unease at violating JPRR. He might be right about this, but I'm not convinced. And I don't think that New-type cases are enough to push me out of JPRR. It isn't just a bare assertion that having "brought x about" is necessary to responsibility, it is relevant to the explanations of why agents are responsible in a wide range of much more ordinary cases, the sorts of cases that ought to inform our theories of responsibility in the first place.

[Saul Smilansky]

Hi,

Thanks, Neal, for re-introducing the issue and for your good comments, to Stephen Kearns and Helen Beebee for their sophisticated and helpful papers, and to everyone for this discussion. This is a new topic, recognized in this (secular) form for less than twenty years, and there have been only something like a dozen papers published altogether, so we are all still trying to figure out elementary things. But I feel that we are clearly making progress.

I agree about 75% with Neal... I don't want to take too long, so I will just focus on what I see as the main problem with Beebee's first (knowledge based) argument, the only argument which Neal was sympathetic with. This will help me to consider Matt's challenge as well. I'll quote first from my reply to Beebe:

"The issue is not only what we may come to know, but under which conditions we are morally permitted to prepunish. On my account of a typical compatibilist situation, the reason why we may punish at t0 is not predictability but that the crime, in a sense, is already there. If a person has formed an intention to commit a crime, and it is determined that this intention will later cause the crime (in a non-deviant way), then we seem to have, long before the crime is actually committed, a punishable desert and responsibility for the crime (and not only for the intention to cause the crime). The passage of time is still required, but merely for the crime to unfold, as determined. For a determinist, even before the crime occurs, nothing morally substantial may be missing, in order for the punishment to be acceptable. Morally, determinism seems to extend the already formed intention into complete causal efficacy and culpable responsibility for the crime, even long before the crime takes place."

As I see it, the question is where to draw the moral line: Beebee's merely-epistemic account is too lax, and lays her open to prepunishing 3 years old (as I explain in my paper replying to her), or, as Mark points out, prepunishing wonderful people while they are still morally perfect (as long as we know that in the future they won't be). Surely we cannot as a rule punish someone when he hasn't a clue why. BUT while Beebee is too lax, Matt in his responsibility condition is too strict, not allowing us by stipulation to see the problem: that in certain situations a reasonable understanding of desert and responsibility can be satisfied, for compatibilists, even before the crime is actually committed. That is what I try to show in the quote above.

[Brian Parks]

Hello everyone, my name is Brian Parks, and this is my first post at the garden!

As a word of caution, I am a grad student in physics, not a philosopher, so I tend to come at these debates from a different angle.

As for my biases, I am a free will skeptic and a moral anti-realist. I think that our intuitive notions of free will and moral responsibility are grounded in a metaphysical conception of ‘self’ that is highly questionable on philosophical grounds and borderline untenable on scientific grounds.

Ultimately, I would say that our ‘intuitions’ about free will and moral responsibility speak to how our brains have evolved, rather than to how the world actually is (at least on a moral level). In that sense, I would argue that the compatibilism-incompatibilism debate amounts to nothing more than a language game that we play amongst ourselves. Reality doesn’t have anything to say about it, at least not in the way that it might say things about theories in physics.

Intuitively, I am drawn to the position of G. Strawson. So when I play the game, I take an incompatibilist position. I think that our evolved inclination to feel strong emotions of praise and blame for other individuals—otherwise privileged as an ‘intuition’ that they are ultimately responsible—does not fit with other ‘intuitions’ that we have, specifically those pertaining to the moral irrelevance of luck. In conflict, I find the latter intuitions more forceful (and would endorse practices of reward and punishment based in a kind of pragmatism).

I think Smilansky’s argument in Analysis is an extremely clever way of expressing the problem that G. Strawson makes in The Basic Argument. Specifically, how can the appropriateness of retributive punishment hinge on a luck-based factor, such as what mental traits an individual naturally inherits, or where in the passage of time an individual happens to be presently located?

What I want to do is propose a scenario for further analysis.

Suppose that W is a non-relativistic world that is deterministic in a specific subset of the states of its trajectory.

(Note, I did not say that W is a deterministic world. Assume W is deterministic in a specific subset of the states of its trajectory, with the nature of the rest of the states of its trajectory unspecified for now.)

(Note, a world W is deterministic in a specific subset S of the states of its trajectory if for any world W* identical in its dynamical laws to W, if any time that a state identical to the first state of S occurs in W*, then every state of S follows in W*, separated by the same amounts of time. This is important for what is to come.)

(So, let’s say that W is deterministic in a specific subset (S1, S2, S3) of its trajectory, with each of these states separated by 3 seconds. This means that for any world W* identical to W, any time the identical state S1* occurs in W*, the equivalent subset (S1*, S2*, S3*) follows in W*, also separated by 3 seconds.)

Now suppose that W is deterministic in the following subset (t0, t1, …, t5) of its trajectory.

t0 John is a 2 month old baby.

t1 John is 25 years old, and has naturally acquired mental nature M, a good moral nature that motivates him to behave compassionately.

t2 John makes a ‘free’ choice, consistent with the desires and beliefs contained in M, to help Steven.

t3 John is 35 years old. By some natural process, John has acquired mental nature M1, a bad mental nature that motivates him to behave selfishly and destructively.

t4 John makes a ‘free’ choice, consistent with the desires and beliefs contained in M1, to kill Steven. John escapes authorities.

t5 John is finally found at the instant prior to his death at age 85, where he can no longer be made to suffer retributively for his crime.

t6 God (who is all powerful but not the author of any moral law, so that we can simulate non-theism) sees what has happened in W*. In order to allow us to punish John, God rewinds the universe back to the same state that it had at t0 (The rewind takes a femtosecond. Time continues to go forward during that femtosecond, but the events proceed rapidly in reverse).

t7 John is a 2 month old baby.

Now, (t7, t8, … t12) follow deterministically.

t8 John is 25 years old, and has naturally acquired mental nature M, a mental nature that motivates him to behave compassionately.

t9 John makes a ‘free’ choice, consistent with the desires and beliefs contained in M, to help Steven.

t10 John is 35 years old. By some natural process, John has acquired mental nature M1, a mental nature that motivates him to behave selfishly and destructively.

t11 John makes a ‘free’ choice, consistent with the desires and beliefs contained in M1, to kill Steven. John escapes authorities.

t12 John is finally found at the instant prior to his death at age 85, where he can no longer be made to suffer retributively for his crime.

t13 John dies.

If possible, it would obviously be permissible on a compatibilist account for us to punish John at t11.5 for what he does at t11. The same is true for punishing him at t4.5 for what he does at t4.

But let’s assume that this isn’t possible in W.

Suppose that it is only possible for us to exact retribution against John between t0-t4 (non-inclusive) and between t7-t11 (non-inclusive), and that doing so will not alter or influence any other aspect of W's trajectory.

Would it be permissibe, on a compatibilist account, to retributively punish John at t10 for what he did at t4?

The compatibilist may want to resist here and say no, but I don’t see much room for resistance. If John’s ‘soul’ carries the responsibility from t4 to t5, so that it would be permissible to punish him at t4.5 for what he did at t4 (if that were possible, which it isn’t in the scenario), then why wouldn’t his ‘soul’ also carry responsibility from t4 to t11, so that it would be permissible to punish him at t10.5 for what he did at t4 (as God has given us the power to do by the rewind)?

Certainly, the rewind at t6 would not erase the fact that he killed someone at t4.

Suppose that the rewind at t6 means that he no longer carries responsibility on his ‘soul’ for the murder. If that is true, then all he has to do at t4 is stay in hiding long enough to get to the rewind at t6, and he’ll be scot-free. But this is absurd.

It gets worse for compatibilists.

If we can punish John at t10.5 for what he did at t4, why not also at t7, i.e., why can't we punish the 2 month old baby? Of course, we have certain emotions (that have well-defined evolutionary origins), that make us not want to do this kind of thing, but if it is the same 'self' in the baby, why would it be immoral?

Now, here comes the real problem that seems to confirm Smilansky's point.

If it would be permissible to retributively punish John at t10 for what he did at t4, why would it not also be permissible to retributively punish him at t10 for what he is going to do at t11?

And if it would be permissible to retributively punish him at t10 for what he is going to do at t11, why would it not also be permissible to retributively punish him at t3 for what he is going to do at t4?

Even worse, why can't we just punish the baby at t0 for the fact that all this drama is going to unfold in the first place!

A compatibilist can assert a principle (say, JPRR mentioned earlier in this thread) that asserts a moral distinction between [being predetermined to do something and having already done it] and [being predetermined to do something and not having already done it], but just asserting such a principle doesn’t give an explanation for *why* the distinction matters, which is what Smilansky seems to be asking in his essay.

If the above distinction is what really makes the difference and leads us to oppose prepunishment, then it should click for us. We should react like “Ahh, that principle makes sense”, in the same way that a deontologist would react to the argument that people should at least be given a chance to do the right thing when the trajectory of their behavior is an as-yet undetermined matter.

But to just assert that the distinction matters because we intuitively oppose prepunishment comes across as ad-hoc and unpersuasive, at least IMHO.

Finally, if this problem also plagues libertarian accounts, as Beebee argues, then let the libertarians answer for it as well. If I remember correctly, Smilansky is a free will skeptic, he believes that ultimate moral responsibility (the kind that justifies intentional retributive punishment as opposed to rehabilitation) is incompatible with determinism and indeterminism. I would agree with that.

Anyways, very interesting essay and conversation, I look forward to more discussions with the gardeners!

[Brian Parks]

Ah, just a quick errata... in t6 I said that God sees what happened in world W*, I meant world W. Also, at some points at the end I talked about punishments at t10, then switched to talking about punishments at t10.5, but they are the same. Sorry about that...

[Brian Parks]

Oh, I just saw one more thing now that I reread my post in full.

One might try to circumvent the scenario by arguing that John's moral responsibility is erased at t6 because God brings Steven back to life. But this response just dodges the example (and even then it doesn't really work... God brings someone back and you're off scot-free? What about the suffering of his family in the intervening years?).

To prevent dodging the example with that type of response, we could just say that God puts a different 'self' inside the body of Steven when he reanimates him, and then modify the details of the definition of 'deterministic in specific subset' to accomodate the change, so that we could still get the effect of t7 --> (t8, t9, t10, t11, t12) even though the state of W at t7 would not be fully identical to t0 (in that there would be a different 'self' in Steven's body.)

[Brian Parks]

"If it would be permissible to retributively punish John at t10 for what he did at t4, why would it not also be permissible to retributively punish him at t10 for what he is going to do at t11?"

I forgot to address why I think this is a problem.

How can it make a difference whether we say we are punishing John at t10 for what he did at t4, versus saying that we are punishing him at t10 for what he will do at t11? The distinction seems rather empty to me.

What does it even mean to say that what is done at t4 and t11 is a *different* crime?

From his perspective, he doesn't have any reason to care whether we say we are punishing him for what he did at t4 or for what he is going to do at t11. Both crimes are equally set in his trajectory, and at the time of punishment, he is equally oblivious to them.

Similarly, whether we punish him at t3 for his crime at t4, or at t10 for his crime at t4, he is in the exact same place in his trajectory when we do it. At the time that we punish him, his trajectory binds him equally to both crimes. And in both cases, the punishment comes a total surprise.

[Ryan Lake]

Brian: I have to confess I'm a little confused by your example. To keep this simple I'll just start with your first question: would it be permissible to punish John at t10 for what he did at t4? It seems to me the answer is very obviously "no", whether you're a compatibilist or not. As Smilansky suggests in the previous comment, as a rule it doesn't make sense to punish a person when he hasn't a clue why he is being punished. And surely John at t10 would have no knowledge of his actions at t4, correct?

Matt: If New is suggesting that we can punish even when conditions of desert and responsibility are not met, then I'm with you (it's been a while since I've read New). I'm certainly in favor of JPRR. I just think the interesting (and challenging) task for the compatibilist is explaining why the conditions for responsibility have not been met prior to the crime, even when the intention has been formed and it is determined that the intentions will cause the crime to occur...

[Ryan Lake]

Brian: Sorry, I just reread your last addition and see you already answered my question! As you say, John at t10 is oblivious to his actions at both t4 and t11. Given that, it's hard to see what would motivate the claim that it is permissible to punish him for his actions at either time.

[Clayton Littlejohn]

Saul wrote:
Beebee's merely-epistemic account is too lax, and lays her open to prepunishing 3 years old (as I explain in my paper replying to her), or, as Mark points out, prepunishing wonderful people while they are still morally perfect (as long as we know that in the future they won't be). Surely we cannot as a rule punish someone when he hasn't a clue why. BUT while Beebee is too lax, Matt in his responsibility condition is too strict, not allowing us by stipulation to see the problem: that in certain situations a reasonable understanding of desert and responsibility can be satisfied, for compatibilists, even before the crime is actually committed. That is what I try to show in the quote above.

Two quick points. I think that Saul is right that Beebee's proposal might be too lax, but I think a proposal I floated in the previous thread might be useful here. Saul is quite right that we cannot punish someone who hasn't a clue why he cannot be punished. That's why I think the compatibilist should appeal to the idea that just punishment will involve two conditions: that the subject deserves to be punished; that the punishment can serve it expressive function, which requires that both parties have epistemic access to the relevant crime. (This rules out the 3 year old case and might, with some finessing, address Mark's problem of prepunishing the virtuous before their falls.) I think this offers a more general response to the problem. It's true that, "in certain situations a reasonable understanding of desert and responsibility can be satisfied, for compatibilists, even before the crime is actually committed", but if just punishment is more than deserved punishment because the justification of punishment depends on its serving its expressive function, this might help show that the compatibilist has a principled objection to prepunishment in those cases where we have a firm intuition that prepunishment is wrong. (Fwiw, I think prepunishment is not always wrong. Not if the subject being prepunished knows that he will perform a specific crime and know he has no intention to act otherwise. In these cases, of course, prepunishment can serve the expressive function.)

[Brian Parks]

Hey Ryan,

Ryan: "I have to confess I'm a little confused by your example."

Sorry about that. My presentation above was admittedly sloppy, so let me try to be a bit more clear.

Justifications for retributive punishment normally appeal to notions of deservingness.

On the standard view, deservingness is a property that an agent possesses that is not concretely manifest anywhere in the agent’s present physical or phenomenological state.

Rather, it is a property that traces backwards in time to some past action that the agent produced. It is a function of the agent’s history.

John, for example, is deserving of punishment right now. His deservingness is not concretely manifest in any feature of his present physical or phenomenological state. It is a function of his history. He killed Steven 4 years ago.

The real problem of prepunishment, in my view, begins with the following question:

If deservingness can trace backwards in time to an agent’s past actions, why can’t it trace forwards in time to an agent’s future actions? If it can be a function of an agent’s history, why can’t it be a function of an agent’s future?

A libertarian might argue that deservingness cannot trace forward because the future trajectory of the universe is not determined. We don’t have a single path, we have different paths, and maybe an infinite number of paths, which might include the path of no path at all.

(Let’s remain agnostic for now about whether this response works.)

A compatibilist who accepts determinism cannot make the same response. The future trajectory of nature is set, determined. There is only one path. If a property like deservingness can trace backwards across that path, why not forwards?

A compatibilist might claim that the future doesn’t exist right now. But the past doesn’t exist right now either (or does it?). If deservingness can be a function of a past that doesn’t exist right now, why not a function of a future that doesn’t exist right now?

Earlier in the thread, Matt asserted a principle called JPRR, that “an individual S can only be justly punished for x if S is responsible for x.”

Presumably, Matt will argue that S can only be responsible for x if x is a present or past action that he performs or has performed. If x is a future action that S will perform, then S cannot be responsible for x.

In response, I would simply ask, why not?

Like everything in this debate, the dialectic is ultimately intractable.

There is nothing to stop a compatibilist from asserting the following as a matter of first principles,

“Deservingness is a property that traces backwards, not forwards, and it exists whether the universe is deterministic or indeterministic.”

I proposed the scenario in order to chip away at that kind of assertion.

Looking at the scenario, I would say that prepunishing John at t3 for his action at t4 is actually more intuitively plausible than postpunishing him at t7 or t8 for his action at t4.

Thus the scenario gives us an example in which it would be more plausible to say that deservingness traces forwards in time than to say that it traces backwards in time.

Ryan: "As Smilansky suggests in the previous comment, as a rule it doesn't make sense to punish a person when he hasn't a clue why he is being punished."

I don’t think that rule works. If I commit a crime, and then take a pill that erases both my memory of the crime and my taking of the pill, will I suddenly stop deserving retributive punishment?

This question, in my opinion, highlights certain commonly neglected issues in the debate about free will and moral responsibility. What is a ‘self’? In what sense does a ‘self’ cause physical behavior? In what sense does a ‘self’ endure through time, so that this ‘self’ that is me right now is the same ‘self’ that existed 10 years go?

Personally, I do not think these issues can be resolved in a way that allows for realism about free will and moral responsibility.

[Ryan Lake]

Brian, thanks. I understand the purpose of your scenario. My confusion (and this is still my sticking point) is the claim that it should be permissible to punish John at t10 for his actions at t4, or t11 (or to punish him at t3 for his actions at t4 or t11, etc). I still don't see what motivates this claim; I don't see any reason why a compatibilist (or anyone for that matter) would find it plausible.

To clarify, it seems to me that John's mental state IS obviously relevant to the question of whether it is acceptable to punish him. As Clayton put it, it seems an essential element of punishment is that it perform an expressive function that cannot be performed if the agent being punished doesn't have epistemic access to the crime.

Granted this is a rough rule that may need some revision and refinement. A case where an agent deliberately blocks his epistemic access to his crimes is tricky. Perhaps we might say that since the agent DID have epistemic to the crime, and since his lack of epistemic access now is a result of his own deliberate actions, he is still responsible (maybe Clayton or Saul have other thoughts on what to say about such a case).

But even given tricky cases like that, it still strikes me as intuitively obvious that we shouldn't punish John at t10 for his actions at t4, any more than we should punish the 3 year old John for his crimes 30 years in the future, or for that matter punish adult John for something he did at age 3. Without epistemic access to one's actions, retributive punishment doesn't appear to make much sense.

[Saul Smilansky]

Clayton,

I am not sure why we need to add on the expressive function of punishment. First, not everyone (even not all retributivists) see it as important. Kant of course thought that we ought to hang the last criminal even if human society was about to fall apart. If a guy deserves prepunishment, then he deserves it, well, because he deserves it. What does expression have to do with anthing? And indeed the expressive function itself seems to depend on the idea that the person punished deserves the punishment. So arguably we need to determine when prepunishment might be just, and then we would know what to express.

[Tamler Sommers]

Saul,

I've always understood the expressive function as a justification for punishment as completely independent of issues relating to retributivism and desert. We punish someone in order to to show society and the world that we do not approve of the actions, that we believe the actions are wrong, whether or not the people committing them deserve blame in the robust retributive sense.

Feinberg's case of the pilot who flies into another country's air space is a nice example. The only way to show the other country that we weren't complicit is to punish the pilot. But we might nevertheless think that the pilot isn't blameworthy. (We might think the act was accidental or maybe a sign of good initiative on the part of the pilot.) By contrast, if you don't punish people for acts--say, paramour killings (another nice Feinberg example)--you're implicitly telling society that you don't find those acts to be wrong. Thus you might want a law that punishes people for killing their wives' lovers even if you don't believe husbands deserves blame for the acts (maybe you think it's out of their control).

So I think Clayton's idea that we want to express disapproval of prepunishing someone could work. And when prepunishing was warranted, as Clayton says, it could still serve that expressive purpose.

More importantly, it looks like you'll be in Boulder for the Rocky Mountain Ethics Congress. See in a couple of weeks! (And I owe you at least a round a beers...) Are any other Gardeners going?

[Clayton Littlejohn]

Saul,

I agree that desert is a necessary condition for just punishment, but I take it if the compatibilist can say that desert is not a sufficient condition for the justification of punishment, the compatibilist can say that they do have a principled objection to prepunishment. Although such punishment might be deserved, desert alone does not justify punishing.

I thought you held that the compatibilists do not have a principled objection to prepunishment. If the compatibilist opts for a view of punishment on which part of the justification needed for punishment is that it serves an expressive function and certain forms of prepunishment cannot serve this function, the compatibilist does have a principled objection to prepunishment.

You might be right to reject views on which a punishment is just only if it serves an expressive function, but it seems you'd have to do that to deny that the compatibilist has a principled way of resisting prepunishment.

[Brian Parks]

Clayton: “I think prepunishment is not always wrong. Not if the subject being prepunished knows that he will perform a specific crime and know he has no intention to act otherwise. In these cases, of course, prepunishment can serve the expressive function.”

What if his epistemic access to the crime gets obstructed pharmacologically, or by some natural process, so that the expressive function of the punishment is no longer fulfilled?

Ryan and I were discussing the case of a criminal who commits a crime and then takes a pill that causes him to forget the crime and to forget taking the pill. He retains the character that led to the crime, but has no epistemic access to it.

Would we still want to retributively punish him for it? I think we would.

This leads me to suspect that the 'expressive function' and the 'epistemic access' that it requires are not the features that do the real work as far as our inclination to oppose prepunishment is concerned. There are situations where we would continue to want to exact retributive punishment even in the conclusive absence of both of those features.

Ryan: “A case where an agent deliberately blocks his epistemic access to his crimes is tricky.”

Well, using the term ‘block’ might suggest that the epistemic access is still there, just in a suppressed form. Let’s use the term ‘eliminate.’ He eliminates his memory of the crime altogether. It is irretrievably deleted from his neural structure.

Why would this kind of case be tricky? If, as you said, it is true that,

“an essential element of punishment is that it perform an expressive function that cannot be performed if the agent being punished doesn't have epistemic access to the crime”,

then a case where he eliminates his memory should not be a tricky case.

When we look at the case, we should react like “well, OF COURSE he shouldn’t be punished, he doesn’t have epistemic access to the crime!”

The fact that we don't react this way suggests that an objection based on epistemic access is not a principled objection to pre-punishment. It’s not the real explanation for why we object to prepunishment.

I think we should abandon trying to resolve the matter from a realist perspective. If we make the question about what types of punishments people ‘really’ deserve, we won’t get anywhere—that question has no fact of the matter.

What has a fact of the matter is the following question. What sorts of factors incline us to want to impose suffering on other people for their actions, and under what conditions are we willing to endorse or follow through on our inclinations?

I think that Smilansky is correct to suggest that indeterminism—the possibility that a person might not perform the action in question—is a major psychological factor that makes us hesitate to want to punish the person for it beforehand.

But as I think more about it, I find myself noticing other inclining factors as well. One crucially important factor seems to be the present state of the individual’s character. Does the individual’s present character fit with the action? I think this is a significant consideration, and I think it explains why we would not want to postpunish John at t7 or t8, but we might actually want to prepunish him at t3 (assuming determinism is true).

As for the question of memory, I do not think it is significant as an inclining factor.

Suppose, for example, that a criminal is born with a bad character, brutally kills someone, and then unknowingly takes a pill that completely erases his memory of the crime. He doesn’t remember the crime, but he is the kind of person that would do it again in a heartbeat if given the opportunity.

Would his lack of epistemic access to his past behavior incline us against retribution? I don’t think so.

Now, suppose that instead of taking a pill to erase his memory, we give him a pill that permanently changes his brain and causes him to have a good character. He becomes compassionate, empathetic, considerate, humble, shy—the nicest, kindest person on the entire planet. He remembers committing the crime, and feels absolutely terrible at the thought of it. He doesn’t understand how he could have done it—but he nonetheless remembers doing it.

Would we still want to make him suffer in retribution? I don’t think so.

I think the reason why we react negatively to the idea of just grabbing someone off the streets and beating the hell out of them for something that they are ‘going to do’ is first because we can’t ever really say that they are going to do it (Smilansky’s point about indeterminism), and also (especially) because it may not reflect their present character.

If the action in question does actually reflect their present character, and we can in fact say with reasonable certainty that they are going to do it as soon as they get the chance, I think we would be inclined to want to prepunish them (if that were our only option), even though they wouldn't be epistemically aware of having committed any crime.

So in that sense, I think Smilansky is right, determinism does have serious implications, because it theoretically removes a psychological consideration that disinclines us towards prepunishment—the idea that the future is still up for grabs.

We have to remember, however, that our psychological inclinations regarding desert—and the neural substrates that underlie those inclinations—are evolved phenomena that have emerged because of their adaptive utility for our species, not because they reflect any kind of objective moral truth about what individuals ‘really’ deserve. Evolution--and the objective universe--don't care about that kind of stuff.

In that sense, the ‘wiser’ part of me is inclined towards the pragmatic view, which would have us repudiate the notion of retributive punishment altogether and take a rehabilitative view that would minimize suffering in the long run. Of course, I openly admit that this is just personal preference, and that I can’t sell it or demonstrate its appropriateness on objective factual or scientific grounds.

Anyways, great discussion!

[Fabio Fang]

I'm not sure if it's just me, but it seems to me that a world in which both libertarianism and eternalism are true is rather implausible (and, a fortiori, so is an indeterministic, eternalist world).

Note that even if eternalism implied determinism (as I'm suggesting), it's not obvious that Smilansky's right when he says that in a deterministic world, but not in the libertarian one, the crime "is already there". That would be the case if determinism implied eternalism, but that's a different implication.

[Saul Smilansky]

Looking forward to seeing Tamler again and other gardeners in Boulder; we could go out for beer and prepunishment.

Clayton, the challenge is for compatibilists to be able to reject prepunishment, as (it seems) libertarians can, in a case such as New proposed (the guy who is planning to overspeed, offers to pay the fine now, and we know we won't be able to catch him later). I don't see that the expressive function of punishment blocks prepunishment in such a case, nor have I seen as yet any other compatibilist way of doing so.

[Alan]

Classical compatibilism has always been about practical values of causal influence related to matters of responsibility, rather than theoretical and more abstract values of Kantian universal justice, retribution after the fact, desert and so on. The New case of the overspeeder is one in which the fine is offered to be paid in light of legalistic rules understood, met, and abided by one who anticipates breaking them. But that flouts the very reason classic compatibilists laid down principles of punishment--to deter future behavior. If someone were so clever as to anticipate breaking the law and offered to pay the relevant fine beforehand, that person could hardly be thought to have grasped the concept of deterrence--rather, they take the law to be a matter of prepayment of anticipated and expected debt--more like taxes than punishment. (Though many would contest the difference!) But the point is not that we expect a payment or prepayment for action--we expect *to prevent the action itself*!! Someone who offered to prepay for an intended future action that we wished to prevent could hardly be said to have grasped the concept of real deterrence of that action. I still feel that undercurrents of desert-entailing responsibility are prejudicing the case against compatibilists here--and compatibilists need not cede ground to that.

[Saul Smilansky]

Alan - your proposal seems to pull compatibilism too much towards some sort of consequentialism. Many compatibilists (myself included, as a partial compatibilist), and probably MOST compatibilists, think that compatibilism can defend the traditional paradigm, wherby control is required for moral responsibility, MR is required for desert, and desert is reqired for just punishment. This does not mean that compatibilists must see punishment as intrinsically good and as not requiring any other (e.g. consequentialist) justification, but they do play the traditional game. Even if I am right that compatibilists have a problem avoiding prepunishment in some cases, that is surely not a sufficient reason to give up on the ambitions of compatibilism.

Fabio - My point is moral, not metaphysical. I claim that morally, once certain conditions are met there is nothing missing for compatibilists to be able to prepunish. So "the crime is already there" should be read ethically.

[Alan]

Thanks for the remarks Saul. Yes, I'd agree that my view of compatibilism does pull hard toward consequentialism (and even in a hard determinist sense, as I'll suggest in a moment). Over the years I've come to see that there are virtues in dropping some of the traditional elements of responsibility. Though said in the context of defending more of a hard determinist position, I'm attracted to some of the same sentiments of Einstein when he remarked in his famous "Credo": "I do not believe in free will. Schopenhauer's words, "Man can do what he wants, but he cannot will what he wills," accompany me in all situations throughout my life and reconcile me with the actions of others, even if they are rather painful to me. This awareness of the lack of free will keeps me from taking myself and my fellow men too seriously as acting and deciding individuals, and from losing my temper." Clearly Einstein rejected free will because he assumed (like Schopenhauer) an incompatibilist view of it; however, his own phrasing above states that a person "can do what he wants"--which is of course a crude account of classic compatibilism. Matters of psychological tracing (e.g.) are important to fine-tuning how even such bare-bones consequentialist responsibility is meted out--but some of the more traditional elements that are emotionally based are probably better left behind, as Einstein says.

[Matt King]

Hi again Saul,

You write: "...once certain conditions are met there is nothing missing for compatibilists to be able to prepunish."

But this can't be what you mean - because even I can agree that a compatibilist has resources to defend prepunishment. So does an incompatibilist - for she would not be forced to accept your principle regarding respect for persons. It's merely open to the incompatibilist to do so.

You mean instead that 'once certain conditions are met nothing is missing to which the compatibilist can appeal so as to *reject* prepunishment'. And here, it seems, we do have some offers on the table (my JPRR, for example). Moreover, if one has a metaphysical view of responsibility, then the metaphysical facts will matter (as I take it Fabio suggests), and so speaking only "ethically" won't alone make the case.

[Brian Parks]

Megan,

In the paper, Eddy et. al. use the following definition of determinism: that the laws of nature and state of the universe at one time entail the state of the universe at later times.

This definition will probably work for the purposes of this debate.

If you are looking for something more philosophically precise, John Earman gives a definition in terms of possible worlds on p. 13, Chapter 2.6 of “A Primer on Determinism”,

“Letting W stand for the collection of all physically possible worlds, that is, possible worlds which satisfy the natural laws obtaining in the actual world, we can define the Laplacian variety of determinism as follows. The world w in W is Laplacian deterministic just in case for any w’ in W, if w and w’ agree at any time, then they agree for all times.”

The term ‘causal’ does not add anything to this definition, other than to suggest that some unspecified type of causation holds in w.

The claim that a person has ‘free will’ implies a privileged mereology of the universe on which there is a distinctly existing ‘self’ with the capacity to ‘cause’ behaviors.

The addition of the term ‘causal’ to the term ‘determinism’, then, should not raise any issues for those who argue in favor of the existence of ‘free will’, because they will eventually have to embrace a specific conception of causation anyway.

We can define the more limited (and more relevant) thesis of neurobiological determinism as follows: for any phenomenological behavior (thinking, feeling, choosing) of a human organism at time t, the behavior is uniquely specified or set by the physical state of the universe (specifically, the physical state of the organism’s brain) at some time t – dt, let dt be any positive number greater than 0.

[Brian Parks]

LOL, dude I am all jacked up this morning. The above post is in wrong thread, sorry moderators and participants...

[Stephen Kearns]

Hello everybody. I'm Stephen Kearns, of "Compatibilism can resist prepunishment" fame. It seems Neal Tognazzi finds Smilansky's response to me decisive. Unsurprisingly, I don't.(Oh, and thanks to Clarissa Millarker for sticking up for me!) I just thought I'd say why I think Smilansky's reply fails. I made two main points. I'll talk about the first point I made in this comment and reply to other in another comment.

My first point, then, was that Smilansky's commonsense reply to prepunishment is available to the compatibilist. In essence, his thought is that we should give the person who will commit the crime the opportunity to change her mind out of respect for her. As many compatibilists believe that the person (even if determined) does have such an opportunity, the compatibilist can appeal to exactly this thought.

Smilansky says that waiting until the crime is committed would be pointless if the person is determined to commit it (and we know this). But that rather depends on what the point of waiting is. We know that the person will commit the crime and that she is able not to (assuming a standard compatibilist view). The point of waiting is to allow her to take this opportunity. This shows her appropriate respect.

Smilansky says this is pointless if we know she won't take this opportunity. But I don't see why this knowledge changes the respect we should show her. Shouldn't we still give her the opportunity to change her mind, given that she really does have this opportunity? Why is showing her this respect pointless? Isn't showing her this respect better than not doing so? It strikes me that Smilansky is thinking that, as there is no pragmatic reason to wait, there is no reason at all. But the reason to wait was never pragmatic, but to show respect for the person.

I am quite confused by Tognazzi's further point that a compatibilist can't allow that the person who is determined to commit the crime has the opportunity to do otherwise. No compatibilist who believes that determined agents can do otherwise will say that they always lack the opportunity to do so. That would be a bizarre view.

[Stephen Kearns]

Sorry, I just realised that I spelt 'Tognazzini' as "Tognazzi" throughout that last comment. Neal--please feel free to now call me simply "Kear" as punishment. Of course, if you do, I will simply have been prepunishing you for doing so.

[Stephen Kearns]

Right, onto the second point I made. I'm afraid it's a bit long.

My claim was that, in the vast majority of cases, prepunishment that doesn't prevent the crime will be such that, (C) if the agent had not been prepunished, the crime would not have been committed. (Tognazzini doesn't quite get my objection right--I appeal to C (or that C might be the case), not simply the idea that the prepunishment might cause the crime).

Furthermore, I pointed out that if we were to concentrate on only those cases in which the person would have committed the crime whether prepunished or not, there are further problems. First, knowing that the person would commit the crime either way would involve knowing more than will actually occur, which will often be a problematic epistemic task. Second, whether one prepunishes the person will often alter her reasons for committing the crime (even rendering the crime innocuous). Third, if the person would commit the crime whether punished or not, then she is more than simply determined to commit the crime. This further element (which I rather dramatically called 'fate' in the paper) could well be just that element that makes a person non-responsible for the actions they perform.

Smilansky's reply is that none of the above problems apply to the original New case. (It is not enough, as Tognazzini suggests it is, just to limit oneself to cases in which the counterfactual C does not apply. This because of the further objections I give above.)

New's case is one in which the person who will commit the crime offers to be prepunished. Smilansky says we know he will (given determinism) commit the crime whether or not he will be prepunished, because he tells us so. Thus my first point (appealing to C) does not apply. And my second point, that it will be hard to know that (C) is false, also does not apply. He also claims that the prepunishment will not affect the reasons for committing the crime, and that (given compatibilism), New's agent can be considered free.

My first point concerning Smilansky's reply is that New's case is a very special case. If compatibilism allows prepunishment in such a case (though I don't believe it does) and not in any others (except very similar cases) this is a very circumscribed conclusion.

My second point is that IF compatibilism must allow prepunishment in New's case, then so must most versions of libertarianism. New's agent freely chooses to commit this crime, and this choice (we may imagine) determines that he commits the crime. On most contemporary versions of libertarianism, the crime is committed freely, as it is determined appropriately by a free choice. Thus if compatibilists are committed to prepunishment in such a case then so are most libertarians.

Thirdly, the fact that the agent tells us he will commit the crime may be enough for us to know that we will commit it, but not that it is both the case if we were to prepunish him he would commit it and if we were not to prepunish him, he would commit it. (In fact, we know that if we were to prepunish him in various extreme ways, he would not commit it).

Fourthly, New's agent's reasons ARE different depending on whether or not he is prepunished. For example, if he is prepunished, he knows that if he does not commit the crime, he was prepunished for nothing. This can act as a reason for him to commit the crime (it may even be moral reason to do so).

Fifthly, compatibilists can quite consistently claim that New's agent does not commit the crime freely if it is really true that he would have committed the crime whether or not he is prepunished. A compatibilist might believe, for instance, that an action is free only if there is a very close possible world in which the action is not performed. Such actions would be very sensitive to conditions as they actually are. If New's agent really would commit the crime whether or not he is prepunished, his action is not very sensitive to conditions as they are. It is thus not free. This compatibilism parallels the kind of libertarianism that says that ALL free actions must be such that, at the time they are performed, they could have not been performed.

I still think then that compatibilism can (quite easily) resist prepunishment. And if it can't then neither can most versions of libertarianism.