Whenever I have had a major life decision to make, it has always struck me as of central – indeed, definitive – importance to think about whether I would regret my decision, if things turned out in one way rather than another.  But I find this a bit puzzling.  I’m going to try out one way of saying why.

We can think of my current expectation for what I will regret in the future as a weighted average of what I will regret in different states of the world, weighted by my confidence that those states of the world will result from my decision.  And we can break down the regrets that I will experience in each state of the world into rational regrets and irrational regrets.

Irrational regrets may matter for what decision I should make – after all, regretting something is one way of feeling bad, and how I feel after making a choice is one among many things that may be relevant for which choice I should make.  But it is hard for me to get my head around the idea that this makes good on the centrality that I feel that considerations about regret ought to have for me, when I am making a major life choice such as where to go to graduate school, which job to take, or who to join with as a life partner.  So let’s screen them out.

The expectation for my rational regret we can then regard as a weighted average of what I will rationally regret in different states of the world.

But it is plausible that it is rational to regret something if and only if you believe that you ought not to have done it.  So this is an expectation about what I will, in the future when I am better informed, believe about what I ought to have done at this time.

Now of course I can see why this would correlate with my expectation about what I ought to do.  But I’m not sure how it would help me to decide what I ought to do.  If my expectation of what I will rationally regret in different worldstates depends on what I will rationally believe that I should have done, in those worldstates, then how does it help me to decide what I ought to do, rather than just being a reflection of what I think I ought to do?


13 Replies to “Expected Regrets

  1. Very quickly: it’s a heuristic. I wish I’d thought of that first, but there’s a 2011 paper by Dan Moller that discusses this. Here’s the basic idea, as I see it: When you adopt the perspective of a future self, the short-term pleasures and pains that can otherwise dominate your deliberative landscape will probably be diminished in force. Instead, long-term considerations, such as those relating to your commitments, will be foregrounded. In this kind of case, the regret heuristic works. It helps to see the good and bad aspects of one’s options in correct proportions. (I’m quoting from an unfinished blog post I wrote for PEA Soup in the spring, discussing the benefits and limitations of the regret heuristic!)

  2. So here is how I think of regret: regret is what you experience when, given how the world turned out to be, you prefer that you would have made another decision. So, for example, suppose you buy a lottery ticket at t=1 but it at t=2 turns out not to be the winning one. Then, given the way this lottery turned out, you regret your choice. Regret then can be stronger or weaker or be absent. For example, if your lottery ticket turns out to be the winning one, you don’t regret your decision of buying the lottery ticket (i.e., your regret-value equals 0). If your lottery ticket does not win, you compare the result with the situation had you not bought a ticket at all. So if the ticket costs $1, your regret-value equals -1. (I am making up these numbers. You should construct some kind of utility scale to compare outcomes).

    Your choice rule seems to be the following: at t=1 choose that action such that the weighted expectation of these regrets (i.e., the weighted expectation value of the feeling associated with one’s choice for each state of nature) at t=2 is minimal. (There are some deep questions here whether that is a reasonable choice rule in the first place, but never mind).

    If this is a correct presentation of your method of deciding, it seems to me that you avoid the conundrum that you pose. The weighted expectation of regrets is not the same thing as actual regret. So it seems perfectly fine to me to choose a course of action where you try to minimize the expectation of regrets at t=1 and then regretting your choice at t=2, given how the world turned out. In other words, you can believe – given your choice rule – that at t=1 you ought to buy a lottery ticket and then at t=2, when you find that it actually is not a winning one, regret your decision to buy the ticket. What determines at t=1 whether you ought to buy a ticket is determined by the *expectation of regret*; what determines at t=2 whether you ought to have bought the ticket is *regret*. No contradiction and a consistent way of determining what you ought to do.

    What might worry you is an intuition that something like the principle of reflection is correct (the principle that say that if at t=1 you believe that at t=2 you will believe that p, you should believe that p at t=1).

  3. NB: There is an old paper in Ethics by Michael Weber on the Allais Paradox where this way of thinking of regret is spelled out. I have written about minimizing expected regret in some paper, but I don’t know anymore where. I would have to look it up if necessary.

  4. Here’s a thought, about which I’m not 100% sure:

    “it is rational to regret something if and only if you believe that you ought not to have done it”. I *think* that might be false. The obligations we are under change over time in virtue of our relationships and other such things, and as a result the status of your later regret hinges on your situation at that time, and not on your situation at the time you made the original decision. Perhaps I ought not have a child at my young age. But if I do, I am in a new relationship which might make it true that I ought not regret having had the child.

    If this is right, it might count *against* using predicted regret to make present decisions, since your present decisions ought not hang on obligations you’ll only later incur.

    As I say, I’m not completely sure about all this. If I recall correctly, it connects with some of Harman’s paper on “I’ll be glad I did it reasoning”. (It also might connect with the claim that desires track reasons, rather than values, which I know you are interested in: reasons but not values can be time-relative, I think. But this is somewhat speculative.)

  5. Related to Alex Gregory’s thought: Sometimes predictable, rational regret doesn’t speak against an action at all.

    Taking an example somewhere in the literature on obligation (I can’t remember where from; I’d actually be grateful if anyone knows):

    A young radical stands to inherit a considerable sum of money when her father dies. She thinks that this money will be very useful to her cause, but worries that it will corrupt her and she’ll keep it all for herself.

    This person might rationally try to set things up legally so that the money will go to the cause, bypassing her. She might anticipate regretting that decision. Given her later evaluative stance, that regret doesn’t seem irrational (unless you think her desire is immorally selfish, and that immorality entails irrationality). But her expected regret gives her, I think, no reason whatsoever to avoid her action.

  6. Also related to Alex’s comment: it seems that there is a class of counterexamples to the idea that rational regret just tells us about what we ought to do, and that is a class of action that changes one’s identity in a way that one oughtn’t pursue (but which will entail no rational regret because of the new identity). So I *think* (on first glance) that Alex’s particular case of a young pregnancy is just one that fits the general idea that morality may sometimes recommend that I not act in a way that will make me the kind of person who is not unhappy about having become that person.

  7. Here’s two things I think about anger.

    1. The feeling of anger toward an action A is (prima facie) evidence that that action is wrong?
    2. Anger about an action A is rational or irrational in virtue of whether it is rational or irrational to believe that A is wrong/bad.

    So, I think there is room for thinking that an emotional state can be evidence for proposition P while also having its rationality supervene on the all-things-considered justification for P.

    If that’s true, then we could say something similar here about regret.

    I fully aware, though, that there is some weird bootstrapping going on here.

  8. Here are a few thoughts about ways in which your expectations about rational regret and your beliefs about what you ought to do might come apart:

    1) It might be true that it is rationally *permitted* to regret something if and only if you (rationally) believe you ought not have done it, without it being true that you are rationally *required* to regret something if you believe you ought not have done it. That view gets some support from cases like the person who has a child at a young age, because it at least allows that they aren’t rationally required to regret having a child. If this thought is correct, your expectations about your future rational regret can come apart from what you think you ought to do, because you may expect that you will, rationally, fail to have some regret that you would also be rationally permitted to have.

    2) Along the lines of earlier comments: what it seems rational to regret are things that you believe you *objectively* ought not have done. But the judgment that motivates you to act, that constitutes your decision, is one about what you *subjectively* ought to do. This also muddies the connection between your expectations about regret and your beliefs about what you ought to do (the ones that you make decisions with, anyway). Of course, we typically think there’s some interesting connection between your expectations about what you objectively ought to do (or about the things that determine what you objectively ought to do) and what you subjectively ought to do, which might repair that link. But maybe not completely.

    3) The principle relating the rationality of regret to beliefs about what one ought to have done says nothing about what *degree* of regret one ought to have. Since ought judgments don’t seem to admit of degrees, you’d have to look for the justification of degree elsewhere. A natural thought is that what degree of regret is rational corresponds to how much weightier the reasons favoring that action are compared to the reasons against it, but I don’t think it’s obvious that that’s the right thing to say. If I have a choice between two options, each of which causes a severe harm to someone (a different someone in each case), and I choose the wrong one, maybe it’s rational to feel a lot of regret about it even if the reasons were close – that is, maybe my regret should respect the full harm I caused this person who ought not have suffered at all, and not just the degree to which their harm outweighed the reasons to inflict it upon them. The principles governing the degree of regret might be pretty complicated.

    Let me clumsily try to bring together some of these thoughts in a case:

    Suppose I am unintentionally pregnant. I am uncertain about two things: whether I am psychologically prepared to be a parent, and whether I will feel miserable after having an abortion. Suppose that given my uncertainty about these things, my non-regret-based reasons for and against keeping the child are tied. But suppose I also know that if I have the child, I will, as allowed by the view suggested in 1), rationally not regret it, no matter how the facts turn out (though if it turns out that I am not psychologically prepared, then it would also be rationally permissible for me to regret it). If I don’t have the child, on the other hand, I know that if the facts turn out so that I am psychologically prepared to be a parent, but I will feel miserable after having an abortion, I will regret having an abortion.

    You might think that in this case, it makes sense to keep the child, because that would minimize my expected rational regret, in a way that isn’t explained by my expectations about what I objectively ought to do. (Someone who had the same expectations about their objective reasons, but who would feel the full permissible regret after keeping the child, might not minimize expected regret this way).

    That seems coherent, maybe?

  9. Bruno Verbeek’s proposal, as I understand it, seems to me the best account of this appealing idea of the rational significance of regret. This proposal is actually a familiar idea. (I discuss the idea in my “Gandalf” paper, but the idea was explored in passing in Savage’s classic work “Foundations of Statistics”, chapter 9.)

    However, it would be better to formulate this idea as identifying a rational choice with one that minimizes the expected degree to which it is appropriate (or correct) to regret one’s choice, rather than as one that minimizes expected rational regret.

    There are two amendments here:
    1. We replace talk of the “rationality” of regret — where “rationality” is an internalist notion, fixed by what is going on in the relevant agent’s mind — with talk of the “appropriateness” or “correctness” of regret — where “correctness” is an externalist notion, fixed by some relation between the agent’s mind and the external world.
    2. We replace talk of one’s actually rationally regretting one’s choice with talk of its being appropriate to regret one’s choice. (This is analogous to the difference between doxastically justified belief and propositionally justified belief.)

    These amendments are important because it doesn’t matter whether one ever actually will become fully informed (or whether one expects ever to become fully informed). After all, one might be making a choice about what should happen after one dies, when (as one knows full well) one will not have any belief about whether or not one should have acted otherwise. But one can still ask: If I made this choice, how likely is it that it would be appropriate to regret the choice?

    All that matters is that each of the states of the world that one regards as possible determines a value for each available choice: the “regret-score” of each choice at that state is the degree to which the value of that choice according to that state falls short of the optimal available value at that state. A rational choice, on this approach, minimizes the expected regret-score.

  10. “But it is plausible that it is rational to regret something if and only if you believe that you ought not to have done it.”

    I don’t think this can be right. Your belief that you ought not to have done something can itself be irrational–and it’s not clear how it can be rational to regret something merely because one has an irrational belief about it.

    A more plausible principle is that it is rational to regret something if and only if, as a matter of fact, you ought not to have done it.

    But in that case there is only one kind of action you can never rationally regret: doing what you categorically ought to do (in a broadly Kantian sense). If you categorically ought to do X, there can be no rational grounds for regretting X. This is why, uncertainty about the future and all, no one can rationally regret doing the truly, categorically right thing to do (hence, Kant’s insistence that a good will gleams like a jewel no matter what else happens).

    Now of course, then the question is what, if anything, a person categorically ought to do. I think Kant had the answer to this question subtly wrong, and give a different answer in my book. But, I think, this is a good way to think about rational regret. One can always find rational reasons to regret decisions that don’t turn out how one likes–except for moral decisions which are in fact categorically right in themselves. They, and only they, can never be rationally regretted.

  11. Thanks, everyone – I’ve been teaching for the last twelve hours so I haven’t been able to keep up. A few thoughts:

    Marcus, I agree with your objection to my formulation, but your principle is not more plausible – it could be that you ought to have done something but that it is irrational for you to believe this. The better fix is that it is rational to regret something if and only if it is rational to believe that you ought not to have done it.

    Ralph, I’m not comfortable replacing rationality with correctness, since I do think that some of the cases where it seems most powerful to me to think in terms of regret turn on the incomplete state of knowledge that I can expect to have in the future – in particular, I think that the presence or absence of counterfactual knowledge can have powerful effects on the rationality of regret, and that many choices involve asymmetries in counterfactual knowledge about what would have happened had we taken the other choice.

    Abelard, the points that you make are the same places where I balked when I was writing out my own post and trying to get the core of the idea out in as few words as possible. I’m worried about the first point, but right now I’m tempted to think that the second point is wrong. The ‘ought’ that we deploy in making a decision is not the subjective ‘ought’ of rationality, but rather is evidence-dependent in a speaker-relative way that makes it relevant to both deliberation and advice (this is a longer topic!).

    Andy: do your views about anger extend to counterfactual anger or imaginative anger? I’m not imagining that I do regret anything so much as thinking about what I would regret, and I wasn’t sure whether this might be different.

    Alex, Ben, Travis: yes, I find some (but not all) examples plausible as well, yet still feel like expectations of regret can be important in some way, particularly in major life decisions, and I find that prima facie puzzling even if I try to screen off this sort of thing.

    Bruno: Yes, the puzzle is supposed to turn on the principle of reflection. I also think that it is key to the examples that interest me that the way we calculate regrets should depend on the information available to me in each worldstate, and allow that I might not have all of the information. For example, rather than only having regret values where I buy the ticket and it doesn’t win, I am interested in regret values for, say, withdrawing a paper from consideration at Mind after 24 months with no response from the editors, when what matters to me is whether, if I had given them longer, it might have been accepted.

    Antti: It may be a helpful heuristic, but I don’t think that could make rational the way that I feel about it whenever I have to make a major life decision. For one, I don’t see how the significance of the life choice should affect how good of a heuristic it is, but I am strongly compelled by the thought that the more important the life choice, the more important it is to think about it through the lens of regret. For another, some of the key cases that I am interested in turn on essentially making payments for counterfactual knowledge – for example, by keeping one’s paper under review at Mind for the 25th month in a row, or by ranking one’s second-preferred option first in the medical residency match process, because that is the only way to find out whether one would have gotten into that program if one had tried. Both of these cases are I think tied to regret.

  12. Hi Mark,

    Not sure if it’s too late to post this, but here are my thoughts:

    You asked how regret would help you decide what you ought to do rather than just being a reflection of what you think you ought to do if your expectations of what you will rationally regret depends on what you will rationally believe that you should have done. First, your expectation of what you will rationally regret is not the same as what you will rationally believe you should have done even if the first depends on the second. Second, although regret itself is backwards looking and what one ought to do is forward looking (another difference), note that having an expectation of what you will rationally regret is both backwards and forward looking, and it is in forming these kinds of expectations that regret would help you decide what you ought to do.

    12 hours of teaching! Yikes! I hope you didn’t have any regrets afterwards and perhaps you will choose not to teach for 12 hours in the future if you did. Also, I am amazed that you can still formulate a coherent sentence let alone a bunch of replies afterwards. Wow!


Comments are closed.