Sunday, November 08, 2009

Portmore on Possibilism

Doug Portmore is inviting comments on his manuscript 'Commonsense Consequentialism'. I've previously discussed earlier chapters here and here. I now want to discuss the concluding chapter, which makes a number of claims with which I strongly disagree.

Firstly, Portmore argues that we need our moral theory to assess the permissibility of act sequences, not just individual acts. He mentions a case in which a patient needs either one of two drugs, but the combination would be lethal. It then seems that the doctor is permitted to give either drug -- each act is permissible considered individually -- but it's impermissible to perform both acts together. Contra Portmore, however, I do not think that this sort of case provides any support for the conclusion that an ethics of individual acts is insufficient.

Instead of shifting our attention to assessing act-sequences, we may simply need to build more information into the token acts that we wish to assess. For example, having already given drug A, the token act of giving drug B is now straightforwardly impermissible. But if we hadn't given the patient drug A, then we would now be faced with the prospect of a different token act of the type "Giving the patient drug B", and this particular token action would be permissible (perhaps even obligatory). So you see there is no need to shift to act-sequences. Portmore rightly recognizes that it's insufficient to assess the permissibility of an unspecific individual act type, but fails to appreciate that this problem is easily solved by instead assessing the permissibility of completely specified individual act tokens.

Not only is the shift to deontic assessment of act-sequences thus lacking in motivation, we may also adduce positive reasons for thinking it incoherent -- as I argue in my old post, 'Only Action is Practical'.

On now to the Actualism-Possibilism dispute. Portmore begins with the standard objection that Actualists posit incompatible obligations. I respond to that charge here. In short, once we reject the misguided application of deontic terms to act-sequences, there is no incompatibility at all. At each moment in time, there is a single thing that the agent ought to do, namely whatever would make things turn out best.

Portmore further objects: "actualism holds both that agents ought to pander to their own vices and that they can limit what's required of them by doing so." This is misleading. Obviously if an agent has the option of ridding himself of a terrible vice (at no comparable cost) then Actualism implies that this is just what he should do. The view just is that an agent should, at each time, perform the act of those available that would actually be for the best. This may well include character development. So Portmore misrepresents Actualism when he later claims that it vindicates Tony's "reasoning on the basis of assumptions [about his future viciousness] that are only true because he has acquiesced to [the] idea that he'll behave badly in the future." If Tony's future bad behaviour is contingent on his current acquiescence, then Actualism will straightforwardly imply that what he should do right now is stop acquiescing!

So, fighting our vices (so far as we are able at any given moment) is great. What Actualism instead rules out is wishful thinking, or imprudently ignoring the likelihood of our well-meaning plans being subsequently (and perhaps disastrously) derailed by our moral flaws -- flaws that we can't currently undermine. When deciding what to do at time t1, we must take into account what we will likely do in the future. We should, of course, seek to influence our future selves in positive directions insofar as we are currently able. But insofar as our future decisions are not under our current control, it would be myopic to just assume in our current deliberations - as possibilists advise - that our future selves will do everything they ought. That way lies disaster.

[Aside: it's worth noting that any moral theory implies that an agent can, by acting wrongly now, limit what they are required to do in future. This is a straightforward consequence of "ought implies can".]

The central flaw in possibilism is that it conflates what is in an agent's control at some time or other with what is under the agent's control at the current moment of decision. For example, Portmore claims that it is "absurd" of actualists to claim that Angry Tony (who will later choose to beat his son to a pulp if he doesn't now let off steam by slapping Valerie) "ought to give into his temptation to hit Valerie even though he could refrain from hitting anyone." This is misleading. "Refraining from hitting anyone" is not among the options that Tony faces at the present moment. His current choice is to slap Valerie or not, and if he does not, he will later face a new choice between beating his son or no-one. Actualists agree that, in assessing this later choice, he ought to hit no-one. (It would indeed be "absurd" to deny this.) But to assess the present choice, we need to hold fixed what is not within the scope of his present control. And if he will actually later choose to beat his son (or if this is overwhelmingly likely) then Tony's present choice is between slapping Valerie or else bringing it about that he will (overwhelmingly likely) beat his son. Since all admit the stipulation that the former outcome is preferable, this is what he should choose. It is not "absurd" -- quite the opposite.

Next, Portmore considers the Correlativity Thesis: S1 has a right against S2 that S2 φ iff S2 has an obligation to S1 to φ.
[I]f CT is true, then Valerie has a right against Tony that he refrain from hitting her only if Tony has an obligation to Valerie to refrain from hitting her. But if actualism is true, then Tony has no such obligation... This is quite implausible, not only in its own right, but also because it makes it unclear why Tony owes restitution to Valerie, which he surely does.

This is no objection to Actualism. Just tweak the case so that Tony won't have any later choice about whether or not to beat his son -- suppose he'll be under a compulsion if he does not now slap Valerie. So now even Possibilists will agree that Tony ought to slap Valerie. Yet he owes her restitution all the same. Obviously something is wrong with (Portmore's interpretation of) CT. Either CT is false, or else Tony can have a (pro tanto) obligation to Valerie to refrain from hitting her even though hitting her is what he ought (all things considered) to do. Either way, the objection to Actualism cannot be maintained.

Finally, Portmore discusses the case "Ten Years Later", whereby an agent buys a car and then ten years later decides to drive while sleepy, and ends up in a fatal accident:
Actualism supposes, contrary to commonsense, that all of those [earlier] decisions were wrong merely because she wouldn't have made the bad decision to continue to drive had she not made the other seemingly good decisions. This just ignores the fact that she could have made all those good decisions without having made the bad decision to continue to drive.

This is an unhelpful case, since one might naturally complain that the agent had no reason to expect a bad outcome from any of her prior decisions. So an Actualist who is also an Expectabilist (privileging the "evidence-relative" spot on the epistemic dimension) is not, in fact, committed to thinking those earlier decisions were wrong.

In addressing the arguments for Actualism, Portmore dismisses the significance of benevolent spectators:
[F]rom the fact that we ought to advise Professor Procrastinate to decline the invitation, it doesn't follow that he ought to decline. These two can come apart, for whereas what we ought to do depends on what's the best that we can do, what he ought to do depends on what's the best that he can do.

It may be unfortunate that Actualists have spoken of the act of advice, since a spectator may have reason to give insincere advice if that would lead to better results than speaking the truth. But that isn't what's happening here, as we can see by skipping straight to the benevolent spectator's preferences. If God (or anyone familiar with Prof P's flaws) sees Prof. Procrastinate accept the invitation to review, she will feel like banging her head against the wall in exasperation with PP's lack of foresight. It is clear, in other words, that by agreeing to review the book, PP is making a bad (regrettable, irrational, whatever) decision.

Portmore goes on to claim that Procrastinate, unlike us spectators, can "control what he'll do if he accepts." That is to say, at a later point in time that will be under Procrastinate's control. But it isn't presently, and the present (as I keep insisting) is when he has to make this decision.

We've seen that Actualism makes claims that can sound absurd, insofar as they are easily mistaken for different claims that Actualists do not actually make. No such misrepresentation is required to bring out the problems with Possibilism. For example, in my Saint-Or-Sinner case, it implies that the agent should accept a deal that will (given their moral imperfection) be almost certain to lead to the torture and deaths of millions of innocents. This seems a bad choice to make in the circumstances, to put it mildly.

17 comments:

  1. Richard,

    You write:

    "So Portmore misrepresents Actualism when he later claims that it vindicates Tony's 'reasoning on the basis of assumptions [about his future viciousness] that are only true because he has acquiesced to [the] idea that he'll behave badly in the future.' If Tony's future bad behaviour is contingent on his current acquiescence, then Actualism will straightforwardly imply that what he should do right now is stop acquiescing!"

    Yes, he *should* stop acquiescing, but he isn't going to. That's the assumption. So how have I misrepresented actualism. If he has decided to acquiesce to his future bad behavior and it is true in virtue of this that he will later beat his son to a pulp if he doesn't now hit Valerie, then actualism implies that he should hit Valerie. And he should, according to actualism, reason just as I've said that he should reason. That is, he should reason as follows: "Since I'm not going to do anything but acquiesce to my future bad behavior and since it's true, in virtue of this, that I'll beat my son to a pulp if I don't hit Valerie, I should hit Valerie."

    This is "reasoning on the basis of assumptions [about his future viciousness] that are only true because he has acquiesced to [the] idea that he'll behave badly in the future."

    -Doug Portmore

    P.S. Why is it that I can't cut-n-paste until I hit preview and then edit and why can't I use html like blockquote in my comments?

    ReplyDelete
  2. Did this comments go through? It was supposed to go first.

    Richard,

    Thanks for these comments. I'll try to respond to them as time permits, proceeding in order. And although I may not find the time to respond to all them (there are quite a few), I'll certainly give them all careful thought and revise the manuscript as appropriate.

    Let's start with act-sequences. In the paper, I argue: "we certainly have moral judgments concerning the deontic statuses of act-sequences, and so any theory that purports to account for our various moral judgments needs to be able to account for our moral judgments concerning act-sequences." For instance, I have the judgment that it would be impermissible for Dr. York to perform an act-sequence that included his giving Zeke M1 at t3 and M2 at t4 -- a lethal combination.

    Do you deny that we make moral judgments concerning act-sequences or that moral theories need to account for such moral judgments? And what about my claim that we can fulfill and surpass imperfect duties only by performing certain act-sequences? Do you deny this?

    By the way, I certainly agree that we needn't appeal to act-sequences to account for intuitions such as (1) that it would be impermissible for Dr. York to give Zeke M2 at t4 if Dr. York has already given Zeke M1 at t3 and (2) that it would be obligatory for Dr. York to give Zeke M2 at t4 if Dr. York has not yet given him any medicine and this is his last opportunity to do so, but I fail to see how this is relevant to my argument. My argument is that we make moral judgments regarding the deontic statuses of act-sequences and that, therefore, we should expect our moral theory to be able to account for these moral judgments and that a moral theory that tells us only what deontic statuses individual acts have can't account for these moral judgments because deontic status doesn't agglomerate over conjunction.

    -Doug Portmore

    ReplyDelete
  3. Richard,

    You also write: "But insofar as our future decisions are not under our current control, it would be myopic to just assume in our current deliberations - as possibilists advise - that our future selves will do everything they ought."

    Possibilists do not assume that our future selves will do everything they ought. Where did you get that impression?

    In any case, you seem to be ignoring the fact that sometimes what we do in the future is under our current control. For instance, suppose that if Tony makes a conscious decision now to turn over a new leaf, he would then in the future succeed in doing what he needs to do to refrain from hitting anyone -- that is, he'll succeed in going to the gym and refraining from beating his son to a pulp. Nevertheless, suppose that Tony makes the conscious decision now not to turn over a new leaf and that, given this, it is true that if he doesn't hit Valerie, he will beat his son to a pulp. Actualism implies, then, that Tony should hit Valerie. Isn't that absurd?

    Just to be clear I define actualism as the following view: Whether or not S ought to ϕ depends on what S would simultaneously and subsequently do if S were to ϕ, and not on what S could simultaneously and subsequently do if S were to ϕ.

    -Doug Portmore

    ReplyDelete
  4. Hi Doug, your original comment showed up on my 'about this blog' post. Thanks for resubmitting it here. Blogger allows some html, like italics and bold, but not the blockquote tag.

    You write: "If he has decided to acquiesce to his future bad behavior and it is true in virtue of this that he will later beat his son to a pulp if he doesn't now hit Valerie, then actualism implies that he should hit Valerie."

    This is misleading. Here's the situation. At the present time t, Tony has the following options: (1) cease to acquiesce to his vicious disposition, (2) do nothing [and hence later beat his son], or (3) hit Valerie. Actualism straightforwardly implies that what Tony ought to do at this time is #1. This remains true even if he will not in fact do #1. It is still the case that #1 is the thing he should do right now, and hence the inferior options (including option #3 hitting Valerie) are not what he should choose to do at this time.

    The core actualist claim is that what an agent ought to do at a time is whatever would be best among their (present) options, given their actual subsequent choices. The possibilist instead claims that what an agent should do now is whatever could be best, i.e. whatever would be best on the (perhaps false) assumption that in future they will always choose as they ought.

    Unfortunately, the extant literature confuses this core dispute with an empty terminological dispute (what Jackson and Pargetter call the 'selection problem') about whether the question "Ought I to phi?" means Is phi-ing the answer to what I should do now? or Between the options of phi-ing and what I would otherwise do, should I phi? Though this is no part of the core Actualist claim, J&P happen to hold that the sentence "Tony ought to hit Valerie" is true, because they interpret it as meaning the latter question: Tony should choose #3 over #2. This latter is a true normative claim, though I wouldn't use the sentence they do to express it -- it strikes me as misleading. Even if Tony should, between the options of #2 and #3, choose #3, this is not the choice he faces. His options at this moment include all of 1-3, and what he should choose, by actualist lights, is #1. (J&P are clear about this, but it has unfortunately confused many of their readers, as was seen in the PEA Soup post I referenced in my linked post.) If Tony is deliberating about whether to do #2 or #3, he is deliberating about the wrong question, as even Actualists can (and should) insist. That is why Actualism does not in fact support the reasoning that you claim it does. At least, it does not support using such reasoning in order to settle the question of how to act at that time.

    ReplyDelete
  5. Hi Doug, your definition of Actualism is unfortunate. (As explained in my previous comment and linked post, it conflates the substantive normative content of Actualism with an accidental semantic claim that J&P happen to endorse along the way.) A better view is:

    Actualism*: Whether or not S ought to ϕ depends on what S would subsequently do if S were to ϕ, and not on what S could subsequently do if S were to ϕ.

    This view gets intertemporal cases like Prof. Procrastinate right (i.e. he ought to decline, if he is not currently able to ensure that he will in future likely fulfill his obligation), and also gets simultaneous cases right: the agent ought to do whatever is best among their present options, which may include (as per #1 in my previous comment) changing their dispositions to act in future.

    ReplyDelete
  6. Richard,

    If it's true that Tony is now deciding not turn over a new leaf, then isn't it also true that Tony would beat his son to a pulp if he were to refrain from hitting Valerie?

    If so, doesn't actualism* imply that Tony ought to hit Valerie?

    You keep on saying that what I'm saying is misleading. But isn't it true that actualism and actualism* imply that Tony ought to hit Valerie?

    -Doug Portmore

    ReplyDelete
  7. Doug, no, Actualism* does not imply any such thing. Again, at time t Tony has the three options I previously called #1, #2, and #3. Actualism* merely implies that the value of #2 [doing nothing] is given by what Tony would actually subsequently choose [namely, beating his son], rather than what he could later choose [e.g. hitting no-one]. So Actualism tells us that option #2 is really bad. (Possibilism incorrectly fails to condemn this as an unwise choice.)

    But, emphatically, actualism* gives us no reason at all to think that he ought to do #3 over #1. What Tony ought to do at time t, on my view, is whichever of the presently available options would turn out best. And that is, quite plainly, option #1. Not #3. No?

    ReplyDelete
  8. You may be misreading Actualism* as saying that whether or not S ought to ϕ depends only on (i) what S would subsequently do if S were to ϕ, and (ii) how this compares to what S would subsequently do if S were not to ϕ. But this is not what Actualism* says.

    The stated thesis doesn't itself specify what alternative options we should compare ϕ-ing against, i.e. the 'selection question'. But my view is that we should "look at all the maximally relevantly specific actions possible at or during that time" (this is also how Jackson and Pargetter say that we should answer the question what an agent ought to do at or during some time).

    ReplyDelete
  9. Richard,

    Deliberation occurs prior to action. My objection concerns how Tony should reason when deliberating about whether to hit Valerie. The point in time at which he is deliberating about whether to hit Valerie at t1 must occur prior to t1.

    So assume that it is now t0 and Tony is deliberating about what to do at t1.

    His mutually exclusive and jointly exhaustive alternatives for t1 are: (i) decide at t1 to turn over a new leaf with respect to his future actions and hit Valerie at t1, (ii) decide at t1 to turn over a new leaf with respect to his future actions and refrain from hitting Valerie at t1, (iii) refrain from deciding at t1 to turn over a new leaf with respect to his future actions and hit Valerie at t1, and (iv) refrain from deciding at t1 to turn over a new leaf with respect to his future actions and refrain from hitting Valerie at t1.

    Assume that if he performs i, ii, or iii, he will refrain from beating his son to a pulp at t4. Assume, though, that if he performs iv, he will beat his son to a pulp at t4.

    Assume also that Tony is at t1 going to refrain from deciding to turn over a new leaf and that Tony knows this. Knowing this, shouldn't Tony conclude at t0 that he will beat his son to a pulp at t4 if he doesn't hit Valerie at t1? And given this conclusion and actualism*, doesn't it follow that he should conclude at t0 that he ought to hit Valerie at t1?

    Of course, I realize that Tony should perform i at t1. But Tony knows that he's not going to. To ignore this fact, would seem to engage in what you call wishful thinking.

    In any case, assume that it's true at t0 that Tony is, at t1, going to refrain from deciding to turn over a new leaf. And assume that if, at t1, he neither decides to turn over a new leaf nor hits Valerie, then he will beat his son to a pulp at t4. Isn't it, then, also true at t0 that Tony would beat his son to a pulp at t4 if he were to refrain from hitting Valerie t1?

    I would really like an answer to this last question.

    ReplyDelete
  10. Yes, it's (stipulated) true that Tony will beat his son to a pulp at t4 if he refrains from hitting Valerie at t1. That is: if he doesn't do your #iii, he will do #iv. So what? It doesn't follow that #iii (hitting Valerie) is his best available option. #ii is. And his best available option is what, I insist, he ought to choose at this time.

    This isn't "wishful thinking". I'm not advising him to choose, between the options of iii and iv, what would actually be worse (namely iv). As an Actualist, I evaluate iv as being a worse option than iii. (Do you disagree with this actualist claim?) But thinking that iii is better than iv does not settle the question of whether Tony should take option iii, because this obviously neglects other -- better -- options that are presently available to him, such as option ii! He won't take ii, and I'd say that iii is the next best after that (on an actualist evaluation of the options). But still, strictly speaking, ii is the option that he really ought to take.

    I hope it is now clear what my view is. It would be enlightening if you could reciprocate by clarifying whether you evaluate option iv as better or worse than option iii. Perhaps you are really an actualist too!

    ReplyDelete
  11. P.S. To be as clear as possible, it's essential to separate what Jackson and Pargetter call the 'evaluative' problem from their 'selection' problem. (See the quote here.) Let's step through your case to illustrate.

    Actualism is a claim about how to evaluate options. That is, to evaluate each of (i) - (iv), we look at what the actual result of each option would be. Hence:

    (i) Valerie slapped (gratuitously), Son safe
    (ii) Valerie safe, Son safe
    (iii) Valerie slapped, Son safe
    (iv) Valerie safe, Son beaten

    Possibilism disputes evaluation (iv). Rather than looking at the actual result of Tony's subsequent choices, we should look at what Tony could in future choose. Hence:

    (iv*) Valerie safe, [Son could be safe]

    Possibilists thus evaluate (iv) as a better choice for Tony to make at this time than (iii). For (iii) essentially involves violence, whereas (iv) is compatible with doing no violence (if only Tony were to choose morally in future -- which he actually won't).

    That's the real dispute.

    Now, you've confused this with a very different dispute, over the selection of options. That is, in asking "Should Tony choose option (iii)?" we might be asking either:

    (1) What should Tony choose out of all the available options i-iv?
    OR
    (2) What should Tony choose out of the two options: iii, and what he would otherwise do, namely iv?

    As already explained, I have no interest in the semantic question of whether the English sentence "Should Tony choose iii?" expresses the first question or the second.

    There is a substantive question here, namely which of these questions a deliberating agent ought to be asking themselves. But then the answer is patently obvious: the agent ought to consider all the available options.

    But regardless, it's vital to note that these 'selection' debates are orthogonal to the question of Actualism vs. Possibilism about how to evaluate individual options in the first place.

    ReplyDelete
  12. Richard,

    I'm afraid that I'm not clear on what your view is. We agree that Tony ought to perform ii. But do you, qua actualist, think that Tony ought to hit Valerie? Do you, qua actualist, think that Tony ought not to hit Valerie?

    When you ask whether I evaluate option iv as better or worse than option iii, are you asking whether things would better if Tony performed iv as opposed to iii? If so, the answer is no. Or are you asking whether Tony has better reason to perform iv as opposed to iii. If so, the answer is yes. I do think that this is a bit of strange question, though. As I see it, actualism and possibilism are two are views about what agents ought to do not views about how to evaluate options.

    ReplyDelete
  13. P.S. I'm asking whether Tony should hit Valerie? That's the question that I want an answer to.

    ReplyDelete
  14. Hi Doug, I think your question is ambiguous, since you haven't specified the selection of relevant alternatives. If you mean something like, "Among all the options available to Tony at this time, is hitting Valerie the thing he should do?" then I would answer 'no'. If you instead mean, "Between the two options of hitting Valerie and what he would otherwise do [namely, iv], does Tony have better reason to hit Valerie?" then I would answer 'yes'. I don't think there's any deep philosophical question about which of these is the correct interpretation of the English question you asked -- that's a matter of semantics, not normative ethics.

    "Or are you asking whether Tony has better reason to perform iv as opposed to iii. If so, the answer is yes."

    Right, that's the question I was after. So you are indeed a possibilist about evaluation. This is where we disagree. Good to know.

    ReplyDelete
  15. "I do think that this is a bit of strange question, though. As I see it, actualism and possibilism are two are views about what agents ought to do not views about how to evaluate options."

    Jackson and Pargetter are quite explicit that Actualism is a view about how to evaluate options. (Again, see the quote here.) And I'll explain below why this is the only substantive normative dispute in this vicinity. But first let us note that it does also have implications for how agents should act in some cases. For example, if we subtract (i) and (ii) from Tony's option set, then Actualists will say that Tony plain ought to do (iii), whereas Possibilists will say that he plain ought to do (iv). This is a real dispute. (As is the Prof. Procrastinate case, on the assumption that he doesn't currently have the option to influence his future behaviour.)

    In other kinds of cases, an apparent dispute about "ought" may be merely verbal. This is so if (as in Wedgwood's case) everyone agrees that:
    (1) option A > B > C, hence
    (2) what the agent ought now (i.e. between all his options) to do is A, and
    (3) if he doesn't do A then he ought (between options B and C) to do B.
    (4) If agent doesn't do B, he will in fact do C.
    (5) Hence, what the agent ought to do between the options of B and what he'd otherwise do (namely C), is to do B.

    These exhaust the normative facts of the situation, as I see things. (Maybe we should add the normative truism that agents should deliberate over all their available options, not just some of them. But I don't think anyone disputes this.)

    Any remaining dispute would appear to merely concern whether the English sentence "Agent ought to do B" expresses the uncontroversially true normative fact in 5, or the uncontroversially false normative proposition (contradicting 2) that between all his options Agent ought to do B.

    This looks to me like a merely verbal dispute. There's no further normative issue to interest us here, is there?

    ReplyDelete
  16. Hi Doug (if you're still reading), you had some questions in your original comment that my above comments don't answer. I've instead responded in a new post: Ethics without Act-Sequences.

    ReplyDelete

Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)

Note: only a member of this blog may post a comment.