Elizabeth Harman

Elizabeth Harman (New York University), "The Mistake in "I'll Be Glad I Did It" Reasoning:  The Significance of Future Desires,” with commentary by Brook Sadler (University of South Florida). Both the paper and the reply can be found here. 

Comments

[Christian Lee]

Dear Elizabeth,

First, I really enjoyed your paper. I think the issue you're discussing is a very important one. I'm not sure though if I understand the proposal quite yet.

You write: "My preference will not be reasonable merely due to features about my future situation that are not true of me now—that is, not merely because of features of my character, features of those I love, and who I love, which cannot justify a preference now."

This is supposed to be an additional consideration in the "I'm Glad I Did It" reasoning. But I'm not yet seeing how it explains why the reasoning is bad.

Intuitively, in the teen pregnancy case, the decision to have a child is unreasonable because even though the agent will be glad she has the child she in fact will have, that does not make her choice a reasonable one. What explains why the reasoning is bad is that even though she will be glad she did, she would be more glad if she did not. What makes one's present preferences about the future rational is not simply that they will be rational in the future, what makes the preferences rational is that not only will they be rational in the future, but that they take into consideration all the relevant options.

So, in the teen pregnancy case, the idea is that the preference to have a child at fourteen is not rational because the mother would have been better off (at the later time) had she had different preferences (at an earlier time). It isn't that her later preferences do not justify the earlier one's, but rather, that the earlier one's are not justified because there are better preferences for her to have then.

Anyway, that seems to me to be a reasonable thing to say. Somehow it seems that you want to say that the later preferences must have a certain content that cannot exist earlier, but can only exist later, and that the failure of the existence of this (de re?) content is what makes the later preferences ineligible to justify the earlier ones. If that is right (I'm kind of guessing), then I don't why this is so. Why must the content be de re to justify the preferences and actions?

[Mark van Roojen]

Elizabeth,

I like this paper. It seems like you have shown that reflection for desires does not work for a certain range of cases.

Since I find it convincing and since you are interested in why reflection for desires fails generally, I'll ask about another sort of case that might or might not involve a distinct sort of failure. Your cases involve situations where the grounds for the values that rationalize the desires change, so that the facts that rationally ground the future desires are not now present. I think here is another sort of counter-example to reflection that has to do with evidence shifting without shifting these underlying facts. (To make it easier to follow, I’ll restate reflection for desires.)

Reflection for Desires: If a person believes that she will come to prefer that p be true, and she believes that she won’t be in a worse epistemic or evaluative position at that time, then she should now prefer that p be true.

Suppose that there is a drug which 999 out of 1000 cures a non-deadly but bothersome ailment. The 1000th time it kills the patient. It seems like I might on that basis believe that if I give the drug I will come to be happy that I did, since in all probability it will cure the patient. (Whether that belief is fully justified given the small chance that it is false seems like it is irrelevant given the statement of reflection; I might in fact believe it on this basis, even if as it turns out I should only form beliefs about what I will desire when I have evidence approaching certainty.) Yet it sure seems to me I should not now prefer giving the drug to the person, since the consequences of the unlikely failure are very bad. (This is a variant on an example of Frank Jackson’s put to another purpose.)

I don’t think that this is the same as the sorts of failures of reflection highlighted by the paper, because those seem to be cases where the underlying facts that make the desires make sense are what shifts between now and the future. In this case, the facts stay the same (all along either this drug is such that it will kill the patient or such that it will cure her), yet our knowledge of those facts does shift. Nor is this a case where our future information will be misleading - it will in fact be completely accurate.

So I guess my question is (1) whether you think this is a failure of reflection for desires, (2) whether this is in fact a different kind of failure than the ones you rightly highlight and formulate suggestions for how to describe, and (3) is there some common explanation for the failures (assuming you think this sort of failure is in fact a failure of reflection)? I won’t be offended if your answer is that this takes you off track from your main interests in reflection so you don’t want to speculate about this other sort of example.

[Elizabeth Harman]

RESPONSE TO BROOK SADLER’S COMMENTS ON MY PAPER:

(I’m responding late so I’m posting this in the comments thread, though I think it’s actually supposed to be posted on the “Papers and Responses” page.. It may get moved there.)

Brook has raised a bunch of interesting and important points about my paper. I won’t be able to address all of them. I’m going to focus on Brook’s questions and proposals about what the attitude of being glad one did something really is.

I say that being glad one did something is having a desire, or a preference, about the past. Brook says that it is impossible to have a desire about the past. But we have desires about the past all the time. Alex, a Red Sox fan, emerges from a long evening meeting and knows that the game has been over for several hours. He hopes that the Red Sox won. He can clearly have a desire about the past; in this case, he doesn’t yet know what happened in the past. Can we only have desires about things we don’t know about? No. I interview for a job and hope to get it, I wish for the job. When I hear that I’m not going to get it, does that attitude disappear? Not at all. I continue to have that very same attitude—I wish I had the job, I desire the job—even though I now know I am not going to get it. One can love someone and wish to be with them, desire to be loved back, even knowing it is never going to happen.

The two attitudes of wishing something hadn’t happened and being glad that something did happen, which are opposing or contrary attitudes, are incredibly common. They are desires about the past, and we feel them all the time. I step in a puddle on my way to work and soak my shoe: I wish that hadn’t happened, I wish I had looked down just before I stepped. I am playing poker and I call a big bet, and win: I am glad I did it, I desire the world as it is to the world as it would have been if I had not called the bet.

While it’s very common to have these attitude about the recent past, they can also be had about the distant past. Thinking back to a mean thing I said to a friend years ago, which hurt her deeply and ultimately destroyed our friendship, I may wish I had not done it. I wish things were now as they would have been if I had not done it.

Brook considers three different ways we might understand the attitude of being glad one did it, and proposes the third as her own view of what this attitude is. I’ll discuss the three ways in turn.

FIRST WAY WE MIGHT UNDERSTAND BEING GLAD ONE DID IT:

Brook first proposes that we might understand the attitude of “being glad I did it” by looking at cases in which the agent doesn’t know what’s going to happen before acting, and then it turns out that things go very well as a result of the agent’s action. She says this isn’t the attitude I mean. But this is the attitude I mean. Nevertheless, these are not the kind of cases I’m talking about, because in such cases one isn’t in a position to know that one will be glad one did it before one acts (or at least, in some such cases one isn’t in such a position; see Mark van Roojen’s comment on this thread and my soon-to-be-posted response to it).

SECOND WAY WE MIGHT UNDERSTAND BEING GLAD ONE DID IT:

Brook next proposes that we might understand the attitude of “being glad I did it” as to do with valuing things and being attached to things.

What is it to value some things? Is it simply to be attached to them? Or is it to see them as valuable?

There are cases where I can predict that if I perform a particular action, I will be glad I did it, because it will change what I value in the sense that it will change what I think is valuable and what I think makes a good life. These cases are incredibly philosophically interesting. But they are not the kind of cases I am talking about. In these cases, Reflection for Desires would not apply because before performing the action I need not think that later, when I’ll be glad I did it, I will be as good at forming preferences as I am now.

The kinds of cases I am discussing in the paper, are cases where what I am attached to changes as a result of my action, without what I think is valuable or what I think makes a good life changing over time. If a woman had a child at the age of fourteen, she may be glad she did it—that is, she may adamantly not wish she hadn’t done it, indeed she feels the opposite of that—because she loves her child and is attached to her child, and does not wish she had the different child she would have had later if she had waited. Nevertheless, she need not feel that her actual child is more valuable, or makes her life better, than the child she would have had. (Indeed, it would be weird of her to believe that.)

THIRD WAY WE MIGHT UNDERSTAND BEING GLAD ONE DID IT:

Brook’s third proposal—the view she actually endorses—is that being glad one did something is simply thinking one made the right choice.

There are two ways you might think you made the right choice.

You might think you made a choice that was reasonable at the time, given what you knew. (We might say it was “subjectively right.”) In the cases I’m talking about, the agents looking back don’t have this attitude, and they shouldn’t have it. For example, the woman who had a child at age 14 doesn’t not think this was a reasonable choice to make at that time, given what information she had then. She thinks it was a terrible choice to make, given what information she had: she knew it was going to seriously disrupt her education and that it would be hard to be such a young mother.

You might think you made a choice that was for the best, or that would have been reasonable at the time, if you had known some facts you didn’t then know (such as facts about how things were going to work out). (We might say it was “objectively right.”) In the cases I’m talking about, the agents looking back don’t have this attitude, and they shouldn’t. There aren’t any further facts available now, looking back, that weren’t available earlier, that would have made the choice the reasonable one.

Finally, Brook suggests that the reason it can be reasonable to be glad one did it while believing that not doing it would have been best is that the question of whether it was a good choice—a reasonable choice, at the time—is just a different question from the question whether it has turned out to be for the best. Of course that’s true, for two reasons that don’t apply to the cases I’m discussing. First, sometimes one is morally required to perform an action that is not for the best (consequentialism is false, I think and many people think); the kinds of cases I’m discussing, however, are cases in which the right choice is the choice with the better outcome. Second, sometimes a choice was reasonable because the agent lacked crucial information about what would happen if it were performed; these are not the kinds of cases I’m discussing.

In the cases I’m talking about, there is no sense in which the agents can reasonably think they made the right choice. Their choices were not justified, and it’s not the case that given some other facts not known when the choice was made, the choices would have been justified.

Indeed, my cases are all cases where one can reasonably be glad one made a choice that one knows was not the right choice.

[Elizabeth Harman]

RESPONSE TO MARK VAN ROOJEN’S COMMENT:

I’m very interested in the questions Mark raises.

I stated Reflection for Desires as follows:

Reflection for Desires: If a person believes that she will come to prefer that p be true, and she believes that she won’t be in a worse epistemic or evaluative position at that time, then she should now prefer that p be true.

But I think it should be stated as follows:

Reflection for Desires: If a person *reasonably believes* that she will come to prefer that p be true, and she *reasonably believes* that she won’t be in a worse epistemic or evaluative position at that time, then she should now prefer that p be true.

In Mark’s case, I am considering giving someone a medicine that has a 99/100 chance of curing a nonfatal disease, and a 1/100 chance of killing the person. Plausibly, I now reasonably believe that the medicine will cure the person, so I now reasonably believe that if I give her the medicine I’ll be glad I did it. But it’s not the case that I should now prefer to give her the medicine. The problem arises because a 1/100 chance of killing someone is much too big a risk to take with someone’s life, but a small enough chance that it doesn’t prevent reasonable belief that the drug will cure her. (Mark restricts his discussion to the first version of Reflection, which doesn’t require the belief to be reasonable. But if that’s the only version the case works against, then a proponent of Reflection could just revise to the second version above.)

This is an interesting case because it’s one where I predict that I’ll get more information, and so that my future self will be in a *better* epistemic situation than me, and it’s information that’s relevant to forming the preference, so it seems that my future self is even *better* at forming a preference on this matter than I am now. So, this almost looks like the kind of case Reflection is built for: a case where your future self is an expert compared to you now, and where your future self should be authoritative.

But, as Mark points out, this case presents a serious problem for Reflection, because this is a case where the preference in question would be reasonable if one had the later information, but is not reasonable now before one has that information, even though one reasonably believes one is going to get that information!

I think that perhaps a proponent of Reflection should revise it as follows to handle this case:

Reflection for Desires: If a person *knows* that she will come to prefer that p be true, and she *knows* that she won’t be in a worse epistemic or evaluative position at that time, then she should now prefer that p be true.

Then the proponent of Reflection should say, as some people have recently said, that (loosely speaking) when the stakes are high it becomes harder to know, and they should say that when I am deciding whether to give the medicine, I do not *know* it will cure my patient.

So that is one important difference between my counterexamples to Reflection for Desires and Mark’s: in my cases, the agents know they’ll be glad they did it. In Mark’s case, there may not be knowledge.

One might object to my revision of Reflection that it should not be about knowledge, because it should be action-guiding and one may not know whether one knows. Two points about this. First, in Mark’s case, I am in a position to know that I don’t know the drug will cure my patient. Second, there’s no reason to think that Reflection is supposed to be an action-guiding principle. It may simply be a descriptive principle about certain conditions under which a preference is reasonable.

[Elizabeth Harman]

RESPONSE TO CHRISTIAN LEE’S COMMENT:

I’ll comment on three suggestions Christian makes.

First, Christian suggests that what’s going on in the teen pregnancy case is that the woman will be “more glad” if she waits and had a child later than if she has a child young, at age 14. This is wrong because the attitude I’m interested in—being glad one did it—is not an attitude involving *happiness* at all. It is an attitude of preference—preferring things as they are to things as they would have been if the other choice had been made. Now, given the nature of love for one’s child, I think it is implausible that the woman who had a child too young will have a *less strong preference* for things as they actually are, with this child, than the preference she would have had, had she waited, for things as they would have been. Given the nature of love for one’s child, the preference for things as they are will be equally strong in the two possibilities.

This point is more striking in a different case. Consider a woman who has a child very young and whose life is seriously wrecked as a result, and predictably so. Suppose she knows that if she hadn’t had a child too young, she never would have had a child (and she knew that at the time she conceived). Suppose that her life is ultimately quite bad, and she would have been much better off not having the child (and thus not having any child). She might know all of this, and yet love her child so much that is glad she made the destructive choice of having the child. If her life really is bad enough, it seems she made the wrong choice in having the child. Nevertheless, her actual feeling of being glad she did it may be much stronger than that feeling would have been in the alternative—because the intensity of being glad she did it that comes out of love for her child may make this attitude stronger than the feeling of being glad one made a choice that in fact made one’s life much better, but didn’t bring a loved one into the world.

Second, Christian says “So, in the teen pregnancy case, the idea is that the preference to have a child at fourteen is not rational because the mother would have been better off (at the later time) had she had different preferences (at an earlier time).” There are indeed some cases in which one’s well-being depends on what preferences one has. But the cases I’m discussing, and the preferences I’m discussing, don’t have that feature. My point is rather that whether the preferences are reasonable isn’t a function of facts about well-being: you may reasonably prefer the actual situation, in which you are worse off than you would have been.

Third, Christian says, “Somehow it seems that you want to say that the later preferences must have a certain content that cannot exist earlier, but can only exist later, and that the failure of the existence of this (de re?) content is what makes the later preferences ineligible to justify the earlier ones.” That’s not my claim at all. Rather, I think the very same preference could be had at the earlier time and would be unreasonable, and could be had at the later time and would be reasonable. In the case of the teenage mother, the preference is “I prefer to conceive[/to have conceived] at age 14.”

[Christian Lee]

Hi Elizabeth,

"Now, given the nature of love for one’s child, I think it is implausible that the woman who had a child too young will have a *less strong preference* for things as they actually are, with this child, than the preference she would have had, had she waited, for things as they would have been. Given the nature of love for one’s child, the preference for things as they are will be equally strong in the two possibilities."

I think this is probably wrong, unless I'm just misunderstanding. I think the woman who had the child young, if she is rational, should have a less strong preference for things as they actually are. Things are much worse for her than they could have been. This is consistent with the love for her child, in the actual and counterfactual situation, being equally strong. She might report this by saying, "I'm prefer to have had this child as much as any child I might later have had. However, I prefer that things had gone different in my life." The point is that the love for her child is not the only fact that is relevant to whether her preferences are rational, other facts include how things would have been better and in what ways if she had waited.

"She might know all of this, and yet love her child so much that is glad she made the destructive choice of having the child."

I see what you're getting at, but I'm not convinced yet. I'm inclined to describe her as being irrational for prefering to have made her life worse off or out of touch with the value she places on her child. Perhaps for her the value for or preference of a life with a child is greater than a life without one, however good it turns out to be.

"There are indeed some cases in which one’s well-being depends on what preferences one has. But the cases I’m discussing, and the preferences I’m discussing, don’t have that feature."

If one's well-being is just a complicated function of the preferences that are satisifed, which is plausible, then it's hard to see how your case could possibly lack the connection to well-being. I know I'm missing something.

"You may reasonably prefer the actual situation, in which you are worse off than you would have been."

We don't share intuitions here. What do you think about the rule of maximizing expected utility? Is it a true constraint on practical rationality?

"Rather, I think the very same preference could be had at the earlier time and would be unreasonable, and could be had at the later time and would be reasonable. In the case of the teenage mother, the preference is “I prefer to conceive[/to have conceived] at age 14."

I'm sorry. I just don't understand what is different at, say 26 than at 13 years in this imagined case. Does the agent acquire a new preference? A new belief? What is it that explains why it is rational later, but not earlier?

[Mark van Roojen]

The move to using knowledge in the formulation of reflection does seem like the way to go.

I'm not sure what to think about citing the high stakes to raise the standards for knowledge, or at least I'm not sure it helps us here. I do have the intuition that I don't know that I'll have the preference later on, and it may well be because of the seriousness of the risk of death. But I'm worried that we could have a structurally similar example where the payoffs have the same relations to one another (that is their desirability stands in the same ratios), but because the disease is much less unpleasant, curing it is not all that preferable to not curing it, whereas the cost in the case where the drug makes things worse is bad, but not on the order of death. That would seem like a lower stakes case which might mean the person could count as knowing what she will prefer. But still you might think that if the bad option is more than 100 times worse than the disease, we should not now prefer taking the drug.