Caspar Hare

Caspar Hare (MIT) is "presenting" his paper "Morphing and Aggregation."  The two invited commentators for his paper are Peter Graham (UMass--Amherst), and Alastair Norcross (Rice University).

Download caspar_hare.pdf
Download grahams_commentary_on_hare.pdf
Download norcrosss_commentary_on_hare.pdf

Comments

[Caspar Hare]

David and Alistair – many thanks for those comments! Let me give you some initial responses as they come to me. I’ll start with Alistair.

Alistair – I argued that broadly ‘person-affecting’ reasoning supports aggregating inter-personally in a certain class of situations. You wonder if that same reasoning supports aggregating inter-personally in a wider class of situations. You have a proposal as to how to extend the reasoning to cover this wider class of situations. And you ask – If it works, what should we make of that? If it doesn’t work, what should we make of that?

I think that these are absolutely the right questions to be asking. And I think that your proposal is very interesting indeed. Before I get to it, a couple of preliminary points.

First, you say that the conclusion that inter-personal aggregation is ‘ethically justified’ in the narrower class of situations is ‘fairly modest’. But note, just to be clear, that the conclusion is not just that inter-personal aggregation is weakly justified in these situations, in the sense that it is permissible to choose to have the healthy child, to save the healthy rather than the sick, the one rather than the many… and so forth. Rather, it is that inter-personal aggregation is strongly justified, in the sense that it is obligatory to do so. There are people who will not find this obvious, particularly in the saving the healthy rather than the sick case.

Note also that, supposing that there are situations in which we are obliged to inflict local harms on people for their own long-term benefit (to vaccinate them against diseases, to kill them to save them from terrible suffering…etc.) the conclusion implies that there are corresponding situations in which we are obliged to inflict local harms on people for other people’s long term benefit (to kill Peter so as to save Patrick from terrible suffering…etc.). Very few people will find this obvious. The utilitarians I know think that it is an unobvious truth, the sort of thing that requires sustained reflection and an appropriately tuned sensibility to understand. Everybody else thinks that it is obviously false.

Second, you wonder if a geneticist might show us that there are no morphing sequences of the kind the argument relies on. I don’t think this is going to turn out to be a problem. What’s the worry? Presumably that the geneticist will say something like this:

“There’s a gene, the A-gene, that makes a big difference to our physical and psychological development. In fact the difference is so big that it is plausible to say, for any person with the gene, that if his mother had conceived a child in all respects like him but without the gene, then he would never have existed. And vice-versa. Furthermore the having or not having of this gene is an all-or-nothing matter. Everybody is either an A-person or a not-A-person. There are no half-A people.”

I have fairly permissive views about personal essence, so I think it unlikely that anything like this will turn out to be true. But even if it is true, we can still construct a morphing sequence from a not-A-person to an A-person (correspondingly physically and psychologically different) in the following way: In the first intermediary world the person is has the A-gene, but that gene is not expressed (think of this as like being ‘switched off’). In the second intermediary world the person has the A-gene, but that gene is switched on for only ten seconds of his life. In the third intermediary world the person has the A-gene, but that gene is switched on for only twenty seconds of his life… and so on.

‘But wait’, you might say ‘there is no mechanism in nature that switches off the A-gene in the way you are describing.’ Maybe that’s true. If so then the intermediary worlds in the sequence are nomologically unlike our own world, and nomologically unlike the worlds at either end of the sequence. But that’s fine. We can have counterparts in worlds that are nomologically unlike our own. Uri Geller did not really bend a spoon with his mind on the Terry Wogan Show, but if Uri Geller had bent a spoon with his mind on the Terry Wogan Show then I would still have existed.

The general point is that metaphysically possible states of affairs are very diverse indeed – at least as diverse as you can imagine. Diverse enough that, for any people in distinct states of affairs, for any non-fragile theory of personal essence, we can construct a morphing series that links them.

And now to your very interesting proposal. The idea is that we might be able to extend the person-affecting reasoning to cover cases where we can bring it about that one person suffers a lot or that many people suffer a little. I don’t want to rule this out. I just didn’t see how to do it when I wrote the paper.

Okay, so the key question is this. We suppose that you are worse off with two broken arms than with one broken leg. And we suppose that there are many intermediary levels of well-being between the broken leg and the two broken arms. Can we then construct a sequence starting with a world in which there’s one person, with a broken leg, and ending with a world in which there are two people, each with a broken arm, such that the person affecting principle (I’ll put it counterpart-theoretically here)

(Dominance)
If everybody in W1 has a counterpart in W2, and vice-versa, and everybody in W1 is better off than his counterpart in W2, then we ought to favor W1 over W2.

entails that we ought to favor each world over its successor?

Your proposal is that we construct a series in which there is a smooth transition from one person to two (increasingly separate hemispheres and so forth). I think I can readily imagine such a series – call it S. But at this stage (I am open to being persuaded otherwise) I don’t quite see how it will do the job. Here’s why. The counterpart relation is 1-1. So it will only be the case that everybody in one world has a counterpart in another world, and vice-versa, if the worlds contain the same number of people. So Dominance only applies to worlds that contain the same numbers of people. But is it the case that every adjacent pair of worlds in S contains the same number of people? No. Because the relation CONTAINS THE SAME NUMBER OF PEOPLE is transitive, and the first and last worlds in S do not contain the same number of people.

What you think is going on here will depend on your views about vagueness. If you are an epistemicist about vagueness then you will think that there is exactly one adjacent pair of worlds in the series – Wn-1, Wn – such that Wn-1 and Wn do not contain the same number of people. Worlds up to Wn-1 contain exactly one person. Wn and beyond contain exactly two people. When does the switch happen? There’s no way of knowing. It’s determined in some mysterious fashion by our linguistic practices. But there’s some world such that Dominance does not entail that we should favor it over it’s successor. So the argument breaks down.

If you reject epistemicism then you will think that there is some (perhaps fuzzy-edged) region in the middle of series S such that it is indeterminate whether adjacent pairs of worlds in the region contain the same number of people. So there are worlds in the middle such that it is indeterminate whether Dominance entails that we should favor them over their successors. So again, the argument breaks down – because it’s manifestly not the case that, if it’s indeterminate whether we ought to favor A over B, and it’s indeterminate whether we ought to favor B over C, then we ought to favor A over C.

I think that, to make the argument work, you need a more liberal Dominance principle than the one I have been working with – one that does not only apply to worlds that contain the same people. Maybe one that applies to worlds that almost-contain the same people? Maybe one that does not invoke a 1-1 counterpart relation? But the worry is that such a principle will longer capture the spirit of person-affecting reasoning. The project here is to use person-affecting reasoning alone to motivate inter-personal aggregation.

[Alastair Norcross]

Hi Caspar,
Thanks for the response. That has helped me a lot in trying to figure out what I am really suggesting! Let me try to push my suggestion a bit more. I am indeed suggesting that the counterpart relation need not be 1-1. But I am not suggesting that the counterpart relation can be 1-2 (or 1-3, and so on). Of course, some counterpart relations obviously can be 1-many, but not necessarily the one that is relevant to the sort of person-affecting reasoning that you are exploring. My suggestion is that the relevant counterpart relation could be 1-(anything-that's-not-determinately-more-or-less-than-one).
The reasoning is similar to what you give to the gender essentialist. If someone objects to your James-to-Jane sequence that no male can have a female counterpart, you reply that, indeed, no determinately male person can have a determinately female counterpart, but that there are cases of gender indeterminacy. A determinately male person could have an indeterminately gendered counterpart. My suggestion is that a determinately singular person could have an indeterminately numbered counterpart. If, in the gender case, we reject epistemicism, the gender essentialist could still try to block your sequence by claiming that it's simply indeterminate whether the indeterminately-gendered person is a counterpart of the determinately male person. If that is the appropriate response to gender indeterminacy, it also blocks the sequence, because it will be indeterminate whether Dominance entails that we should favor the world with the better off indeterminately gendered person over the world with the worse off determinately male person. I don't know whether this is the appropriate response to the postulation of gender indeterminacy (I suspect not), but it seems to me that the appropriateness of the response is pretty much on the same footing as the corresponding response to number indeterminacy. That was the point of my putting it as a dilemma.
If number indeterminacy about persons is possible, and the more I think about it the more I think that it is, that will certainly affect the nature of the counterpart relation used in person-affecting reasoning. But the same is true of gender indeterminacy. Suppose that there is a possible world in which I am neither determinately male nor determinately female. Now suppose that there is a possible world in which I am neither determinately one person nor determinately two. Reflection on split brain and multiple personality cases makes the latter possibility just as intuitively plausible as the former. Besides, as you point out, the worlds in the morphing sequence don't all have to be even nomologically possible.
So, my suggestion is that, whatever the 'spirit' of person-affecting reasoning is, it's equally captured/violated by a Dominance principle that uses a counterpart relation that can go from determinately male to indeterminately gendered, as a Dominance principle that uses a counterpart relation that can go from determinately singular to indeterminately numbered.

[Agustin Rayo]

It is perhaps worth mentioning that on a standard characterization of 'determinately', 'I am neither determinately one person nor determinately two' counts as a logical falsehood. This needn't settle the issue at hand, since one might complain that standard characterizations of 'determinately' prejudge the issue. But it does suggest that thinking about this sort of case properly calls for non-standard machinery.

[Caspar Hare]

Agustin -- why is that?

[Caspar Hare]

Peter – Many thanks for your comments. You have picked up on an important subtlety in the way the Principle of Irrelevant Alternatives works. I talk about it a bit in ‘Rationality and the Needy’, but I was glossing over it in this paper.

How to capture the idea behind the principle? The first thought you might have is to put it in terms of preferences (this is what is sometimes known as ‘contraction consistency’):

For complete states of affairs S1, S2, S3, if I am rational and I believe that S1, S2, S3 may be actual, and I prefer that S1, rather than S2, be actual, then I will retain that preference upon discovering that S3 is not actual.

But this is too weak to generate many substantive rational constraints on intentional behavior. Suppose that, when options A,B,C, are open to me I take A, but when I discover that only options A,B are open to me I take B. If my actions reflect my preferences between complete states of affairs, then you would think that I must be violating contraction consistency. But it may be that I am not. It may be that the discovery does not lead me to change my preferences between complete states of affairs at all. There are four complete states of affairs at issue.

S1 – in which options A,B,C are open to me and I take A.
S2 – in which options A,B,C are open to me and I take B.

S3 – in which options A,B are open to me and I take A.
S4 – in which options A,B are open to me and I take B.

It may be that, before and after the discovery, I prefer S1 to S2, and S4 to S3.

So, if the principle is to generate substantive rational constraints on intentional behavior, it must be supplemented in some way. How to supplement it is a tricky matter. We don’t want to say that it is always irrational to have preference-patterns as above, because sometimes it is manifestly rational. Sen has some examples. Your worm-eating example is another. The trick is to distinguish the times when it is okay from the times when it isn’t okay. A natural thought, along the lines of what you suggested, is this:

Intentional switching behavior (from A,B,C – choose A, from A,B – choose B) is irrational unless it comes about as a result of my finding S1 intrinsically preferable to S2, and S3 intrinsically preferable to S4, and something about what I value explains these preferences.

This accounts for what is going on in your case. The complete states of affairs at issue are:

S1 – in which the options Not-Eating-and-Giving, Not-Eating-and-Not-Giving, and Eating are available to me, and I give, keeping my promise.

S2 – in which the options Not-Eating-and Giving, Not-Eating-and-Not-Giving, and Eating are available to me, and I don’t give, breaking my promise.

S3 – in which the options Not-Eating-and-Giving, and Not-Eating-and-Not-Giving, are available to me, and I give.

S4 – in which the options Not-Eating-and Giving, and Not-Eating-and-Not-Giving, and I don’t give.

Why do I give in the three-option case, but not in the two-option case? Because I find S1 intrinsically preferable to S2, but not S4 to S3. Why do I have these preferences? Because I value keeping my promises. This doesn’t make me irrational.

Quite right.

Which brings us to the case at hand, where Charlotte can bring James, Janus or Jane into existence by pressing buttons B1, B2, B3. Here the relevant states of affairs are:

S1 – in which Charlotte can press B1,B2,B3, and she presses B3.
S2 – in which Charlotte can press B1,B2,B3, and she presses B1.

S3 – in which Charlotte can press B1,B3, and she presses B3.
S4 – in which Charlotte can press B1,B3, and I press B1.

Your proposal, as I understand it, is that, if Charlotte cares about things being better for any particular baby she has (more precisely: if her preferences conform to the Dominance principle), it makes sense for her to prefer S1 to S2, but not S3 to S4, because in S1 she could have done better for her child, but in S3 she could not have done better for her child. Doing better for her child is something she cares about. So switching behavior on her part need not be irrational.

But notice that caring about things being better for any particular baby you have is not the same as caring about whether you make things better for your particular baby. Charlotte’s preferences conform to Dominance, so she prefers a state of affairs in which she has baby Q, with well-being n, and she could have done better for him (WAY better, even), over a state of affairs in which she has baby Q, with well-being n—dx, and she could not have done better for him.

Furthermore, notice that, even if there is this second thing that she cares about (her making things better for her baby) it is not only her caring about this second thing that explains why she presses B3 in the three option case. As I argued, if she is rational and her preferences conform to Dominance and she has no further preferences, then she will press B3 in the three option case. So she can’t say ‘I press B3 in the three option case, because S2 has a feature (my failing to make things better for my child) that I find undesirable. S4 does not have this feature, so it’s okay for me not to press B3 in the two option case.’ She would press B3 in the three option case even if she did not find the feature undesirable.

[Alastair Norcross]

Augustin, does the same hold for "I am neither determinately male nor determinately female"?

[Caspar Hare]

Alistair – Agustin is on walkabout in Australia, so let me guess at what he meant.

Suppose we have a split-brain sort of case, where it is plausibly indeterminate whether there is one person there or two, and suppose the person says (the two people say) ‘I’. Who is being referred to? A natural thing to say is that it is indeterminate whether one person has spoken, referring to himself, or two people have spoken, each referring to themselves.

Suppose, then, that the one person says (the two people say) ‘I am two people’. Thought about one way there’s one person there, saying something false (the one person is not two people). Thought about the other way there are two people there, each saying something false (neither of them is two people). Either way, everything that was said was false. There’s no way of precisifying the vague concept of a person that makes an utterance of ‘I am two people’ come out true. So (on the super-valuationist treatment of indeterminacy) ‘it is indeterminate whether I am one person or I am two people’ comes out false in this, and in all, cases.

The same does not apply to ‘it is indeterminate whether I am male or female’. Or to ‘it is indeterminate whether I am bald or hirsute’… and so forth. Utterances of those sentences may be true, on the standard reading.

I don’t think that this is a deep problem for your idea. If you find yourself in the split-brain predicament, you can say, instead: ‘it is indeterminate whether one person is uttering these words or two people are uttering these words.’ That comes out true, on the standard reading.

But, for good reasons or bad, some people certainly are going to quibble with the suggestion that the counterpart relation can obtain between a person and an entity such that it is indeterminate whether it is one person or two. Some people will say that there are no entities such that it is indeterminate whether they are one person or two. Think about the split-brain case one way and you have one person there. Think about it the other way and you have two people there. But when two people exist there is no third entity that is the two people, or is composed of the two people, so neither way do you have an entity that is (or is composed of) two people there. So ‘there is an entity such that it is indeterminate whether it is one person or two’ is always false.

There are no similar grounds for worry about ‘there is an entity such that it is indeterminate whether it is male or female’ or ‘there is an entity such that it is indeterminate whether it is bald or hirsute’, obviously.

And other people will say that, even if we concede that there may be entities such that it is indeterminate whether they are one person or (the fusion / composition of) two people, such entities are not counterpart-related to entities that are determinately people, because:

(EDP)
Entities that are determinately-people are essentially determinately-people.

You argued that saying this is just like saying:

(EDM)
Entities that are determinately-male are essentially determinately-male.

Which shows that the one-to-two morphing argument is in the same boat as the one-to-one morphing argument. If EDP and EDM hold then both arguments work. If they don’t then both arguments fail.

But there are famous reasons for accepting EDP that are not reasons for accepting EDM (indeed I don’t know of anybody who accepts EDM, and I don’t see any reasons at all to accept it.) EDP is supported by the idea (to put it in the Wiggins-way) that PERSON is a substance-sortal, not a phase-sortal. And there are arguments, to do with reference, conceivability, ontological economy that are supposed to convince us that this is so. (Eric Olson’s book contains the best discussion of these arguments that I know of).

I don’t find these arguments particularly convincing, but the point is that the matter is controversial. So, while you don’t need to make any controversial assumptions about the metaphysics of persons to run the one-to-one morphing argument, you do to run the one-to-two morphing argument. And that’s unfortunate. It would be nice to rise above quibbles from the metaphysicians.

Maybe you can construct a different, quibble-free kind of morphing argument to the conclusion you want. Here’s a way of illustrating the central thought. Let’s suppose that it is worse for any particular person if he suffers level 9 pain for two hours than if he suffers level 10 pain for one hour. Now take two complete states of affairs in which two people exist for four hours.

S1

During hour 1: A is fine and B is fine
During hour 2: A suffers level 10 pain and B is fine
During hour 3: A suffers level 10 pain and B is fine
During hour 4: A is fine and B is fine

S2

During hour 1: C suffers level 9 pain and D is fine
During hour 2: C suffers level 9 pain and D is fine
During hour 3: C is fine and D suffers level 9 pain
During hour 4: C is fine and D suffers level 9 pain

By the straight morphing argument you will, if you are decent and rational, favor S1FirstHalf (a complete state of affairs just like S1, but with the last two hours lopped off) over S2FirstHalf. And you will, if you are decent and rational, favor S1SecondHalf (just like S1, but with the first two hours lopped off) over S2SecondHalf.

Now suppose:

(Compositionality)
For complete states of affairs I,J,K,L, if you are rational and you favor I over J and K over L, then you will favor the composite state of affairs I+K over the composite state of affairs J+L.

It follows that, if you are decent and rational, you will favor S1 over S2. And so (by careful repetition of the argument) you will favor one death over many headaches and so forth.

This argument doesn’t rely on any controversial metaphysics, or on any intuitions about impersonal betterness, just the claims that I have been calling Personal Dominance (if you are decent you will favor one state of affairs over another if it is better for everybody), Transitivity (if you are rational then your preferences between complete states of affairs are transitive) and Compositionality. But Compositionality is a little bit dubious. Hmm…