A Curious Miscellany of Items Philosophical, Historical, and Literary

Manus haec inimica tyrannis.

Tuesday, September 7, 2010

The Incredible Shrinking Self

I proposed in the post before last to discuss my objections to the “multiple selves” approach to human agency. To refresh your memory, this is roughly the idea that instead of viewing the human agent as a single overarching self that makes choices from some set of feasible options and that retains its identity through time, the agent should rather be seen as an indefinite number of separate selves existing at different times. Some of these selves may have preferences and projects which happen to overlap, but insofar as they do not, this is to be explained by the fact that they are different persons in some relevant sense.

The idea that each of us is but a series of temporally separate selves is supposedly congruent with phenomena like time inconsistency: we tend to rate a good as better the nearer it is to us in time, while rating a prospective bad as less bad the further off in the future it is. An example of time inconsistency would be a case where I wish at time 1 to quit smoking, but at time 2 I wish for nothing more than a cigarette. Thus, my self at time 2 may end up doing something (having a cigarette) that is directly against the interests of my self at time 1. A well-known example of a certain kind of time inconsistency is what is called hyperbolic discounting (the phenomenon is called “hyperbolic” discounting because, when plotted on a graph, the shape of the discount function is a hyperbola).

Put simply, hyperbolic discounting occurs when two goods of equal utility are not valued equally when they offer themselves at different times. For example, it is often the case that an agent would prefer to have $10 now, rather than $30 a year from now. This represents a huge discounting rate: here money loses 200 percent of its value in the space of a year. In reality, the discount rate in this example is quite moderate. George Ainslie notes that “both normal and impulsive human subjects, reporting how long they would wait to get double a hypothetical prize, usually generate answers reflecting annual discount rates in the billions, trillions, or quadrillions of per cent; and a large proportion of adult human subjects who could wait 3 days to receive 25 per cent higher subject pay chose not to do so, a choice which represents a rejection of an annual 5 billion per cent interest rate” (Ainslie 138-139).

With such steep discounting, it is a wonder that we ever manage to delay immediate gratification. As Ainslie sardonically asks, “How may we reconcile this finding with the widespread savings ethic? What would even give primitive farmers a sufficient savings tendency to get through the winter?”

Parfit’s Present Aim Theory.

In the course of presenting the “multiple selves” theory, I had dropped the name of the philosopher Derek Parfit, and I’d like to use his approach to the subject as a starting point for my critique.

Parfit uses the phenomenon of time inconsistency and hyperbolic discounting to argue for his Present Aim Theory of rationality, which says roughly the following: Hyperbolic discounting shows that there is a sense in which I am not necessarily one self, but many sequential selves. The idea that I am one unified self stretching across time is more fiction than reality. Thus, where I find my later self having different interests and evaluations from my earlier self, it is not necessarily the earlier self whose interests must take priority. If indeed we take seriously such a metaphysic of many successive selves, then we must seemingly treat each such self with the same respect as we would any other moral agent. Therefore, if, for instance, at time 1 I want (all things considered) to quit smoking, but at time 2 I want (all things considered) a cigarette, then at time 2 my desire to smoke must be given just as much weight — more weight in fact — than my previous desire to quit. My earlier self’s desire for a smoke-free life was a different person’s desire from my later self’s desire to smoke.

There is both a metaphysical and an ethical aspect to Parfit’s Present Aim Theory. There is the “successive selves” metaphysic, and following from that is Parfit’s utilitarian ethic, which says that since each successive self is a separate person, its preferences are worthy of moral consideration just as any person’s preferences are. We do the most overall good, from a utilitarian perspective, by satisfying as many desires as we can at any given time. We should not make temporal (or other) distinctions. This is, according to him, the best way of aggregating preferences. To use the terminology of my discussion of Elster’s Ulysses and the Sirens, the Present Aim Theory is locally-maximizing rather than globally-maximizing.

Parfit’s approach is to accept time inconsistency as natural and not necessarily pathological, and to take it to its logical conclusion. That conclusion is counterintuitive, and Parfit admits as much, but he’d say that given time and education we can overcome the prejudice that would have us believe that there is a single self, or that some selves’ desires are normatively prior to other selves’ desires. Luckily, there is more than one reason for us not to accept either a successive selves metaphysic or a Present Aim Theory.

“Thin” versus “Thick” Selves.

First, we can question the implicit claim that one or two differing desires at different times are enough to make for different selves. Without delving too much into the problem of personal identity, we can say that though at one time I want to quit smoking and at another I want to smoke, both “selves” still share too much in terms of memories, interests, and desires for us to really consider them distinct selves. If all that changes in me is a desire to smoke, is that enough to form the basis of a new person? Or is what we mean by a “self” or “person” much thicker than this? To accept the successive selves metaphysic is to accept the possibility that a person can be identified with a single desire, such as a desire for a cigarette.

Is it really the case that by according rights and respect to an agent we intend to accord them to something so thin, so minimal? If so, we would have to accord such rights and respect to dogs and cats, and not only such, but also to temporally successive dogs and cats. Anything, in fact, capable of desire, and capable of changes in desire, would be due such moral consideration. Of course, Parfit might be willing to “bite the bullet” and accept such consequences. However, I believe that most of us have something thicker, more substantive in mind when we think of a “person” and a “self”.

(Incidentally, much the same argument as mine has been deployed to deny full personhood to the separate identities in many cases of multiple personality disorder. The separate “identities” are usually too thin to constitute individual persons. Instead they should be viewed as instances of personality fragmentation. They are parts of a self, rather than full and autonomous selves. As such, their psychotherapeutic “exorcism” is not morally equivalent to homicide. Indeed, the therapist doesn’t see herself as getting rid of these separate identities so much as putting them back together into a single whole. Like Humpty Dumpty. See Radden 37ff.)

Interself Epistemology.

Second, there is a notable difference between separate agents on the one hand, and the separate “selves” of Parfit’s scheme on the other. Agents’ access to other agents’ minds is highly restricted, whereas a later self would seem to have privileged access to an earlier self’s mind. This is an expression of what Jennifer Radden has, in another context, referred to as an “interself epistemology” (Radden 47).

The relation seems to be asymmetric: a later self has access to an earlier self’s mind, but the reverse is not the case. However, this may be somewhat overstating the point: although later Jack has privileged knowledge of earlier Jack’s mind in a way that Jill does not, barring self-deception or willful ignorance, it does not take much imagination to see that earlier Jack likely has more insight into what may be going on in a potential later Jack’s head than he has into what goes on in Jill’s.

Causal Proximity of Selves.

Third, it should be evident that a later self is causally related to an earlier self, in a stronger sense than can be the case with separate agents proper. One agent can causally influence another, just not as thoroughly as she can causally influence her later self. I can make or persuade someone to do what I want them to do on occasion, but it is less simple to make them be someone I want them to be. Even parents can hope for only limited success in causally influencing their children’s identities.

We can go further. Assuming my earlier self is to be considered an autonomous agent, the later self is not just causally related to the earlier one, but he is in a sense a creation of him. Thanks to path-dependency, later Jack has been formed by the previous choices of earlier Jack. By contrast, we cannot say that Jack is formed by the previous choices of Jill, or at least not to nearly the same degree.

The scheme of Parfitian successive selves seems to rely upon a mistaken “snapshot” view of the self, where one takes a temporal slice or snapshot of the agent and accords it full personological or moral standing, while losing sight of the fact that the snapshot is embedded in the larger panorama that is the life course of the agent. Such a snapshot is necessarily a distorted one, lacking all connection to the other slices of agency. It ignores the continuity and basic integrity of character, finding instead mere fragments.

Pride, Shame, and Guilt.

Fourth, successive “selves” usually have an emotional investment in each other that is of a different nature from the emotional investment — if any — between different persons. I do not, under normal circumstances, feel shame or guilt at the actions of others the way I might feel them towards the actions of other selves. Instead, it is more proper to say that I can feel embarrassment or anger at the actions of others. Shame and guilt, on the other hand, are self-directed. Such self-directed emotions can be called emotions of self-assessment.

(Pride too is an emotion of self-assessment. This may seem strange at first glance, given that I may, for example, feel proud of the actions of other persons. But as Hume pointed out, such pride has, at bottom, some reference to myself. Thus, I am proud of my daughter’s stellar report card because her achievement at school is perceived as somehow reflecting on my own self-worth. This is not to say that I do not at the same time feel simple joy at her success, irrespective of its effect on myself; but pride and joy are not the same emotions. Pride is self-directed while joy may be other-directed.)

All of these observations serve as refutation of a Parfitian metaphysic of successive selves. The relation between my temporal selves is not relevantly similar to the relation between different persons proper. Thus, we can safely reject the Present Aim Theory of morals that Parfit claims follows from such a metaphysic. Rejection of the Present Aim Theory is not the only normative consequence attending this asymmetry. The existence of emotions of self-assessment and the interest an agent takes in her other temporal selves generate duties of a quite different nature from whatever duties follow from agents’ status as different persons from oneself. The nature of moral praise and blame is different in the two cases: I can direct approval, anger, disappointment, and so on, at myself and others; but pride, shame, and guilt can only really be directed towards myself.

Besides helping to refute the metaphysics of successive selves and any Present Aim Theory following from it, emotions of self-assessment constitute a form of endogenous self-binding that can help an agent overcome phenomena like hyperbolic discounting that may work against an agent’s longer-term interests. (By way of reiteration, in the Parfitian scheme, there is, strictly speaking, no such thing as an agent’s “longer-term interest”. There are only fleeting present interests of multiple agents who happen to have a time-share arrangement over a body.)

For example, the tendency to feel ashamed of precipitant discounting behaviour may contribute as a corrective to such behaviour. And pride felt in the achievement of continence may serve as a reinforcement of such continent behaviour. At the very least, foreseeing that a course of action is likely to result in avoidable and unpleasant feelings of shame can be a consideration affecting my evaluation of a present good vis-à-vis a future one, making the discount curve that much less steep.

Aims Outrun Agency.

Some of what is contained in the above refutation of Parfit was already developed by Christine Korsgaard in her 1989 paper “Personal Identity and the Unity of Agency”. Korsgaard rightly notes that as agents, many if not most of the projects we choose and pursue take time to achieve: “Some of the things we do are intelligible only in the context of projects that extend over long periods. This is especially true of the pursuit of our ultimate ends. In choosing our careers, and pursuing our friendships and family lives, we both presuppose and construct a continuity of identity and agency” (Korsgaard 113). A Parfitian present self’s aims often extend beyond the “life” of the present self whose aims they are. Thus, the “present aims” in the Present Aim Theory are more fiction than reality. Aims outrun agency.

It is simply not intelligible, or at the very least foolishly optimistic, for a self whose “shelf life” does not extend beyond perhaps the next minute or so, to set itself ends achievable only weeks, months, or years into the future. It would be as if I were to make plans on the expectation that I would live to be a thousand years old. On the most optimistic forecast I could live to perhaps the age of 120 years. Therefore, my ends should be tailored to the time allotted. As a matter of fact, the expectable human lifespan is enough time to achieve most of the ends we choose to pursue, fate and circumstances willing. But a Parfitian self may not be allotted enough time to expect any end to be fulfilled; at the very next moment it is at least conceivable that a future self will throw out an end in the pursuit of a different end. A life cannot be constructed from such disposable material. And the problem is not the change of mind per se, but the fact that because the new mind belongs to what is morally equivalent to a different person, the new mind must have its way, even at the expense of interests of selves who no longer exist (or who do not yet exist). We’re not just talking about a change of mind; we’re talking about a change of minds.

According to Korsgaard, the unity of the self is not so much metaphysical as practical. Such unity derives in part from the brute practical need to eliminate conflicts in motivation, in order to serve ends that extend beyond the current moment (p. 110). The Parfitian view identifies the agent with his occurrent set of desires at a given time. Korsgaard on the other hand locates agency in the choosing from among such a conflicting set of desires according to one’s reasons for choosing:

“When you deliberate, it is as if there were something over and above all your desires, something that is you, and that chooses which one to act on. The idea that you choose among your conflicting desires, rather than just waiting to see which one wins, suggests that you have reasons for or against acting on them. And it is these reasons, rather than the desires themselves, which are expressive of your will. The strength of a desire may be counted by you as a reason for acting on it; but this is different from its simply winning.” (p. 111)

This is not to say that it is never the case that a desire simply wins out. We are not always as reflective as Korsgaard’s account would have us believe. However, note that even when this is the case, we may still find ourselves subject to those emotions of self-assessment discussed above. I may regret that a desire has won out and feel guilty for pursuing it. But such guilt would be unintelligible if different desires were “on all fours” with each other, where the desire that wins just wins, end of story. There is no room here for the intelligibility of guilt or shame. After all, what should I care if a previous self (who is not me) chose to pursue a desire that my present self no longer shared? It would be as if I were to feel guilty for someone else who lights up a cigarette, just because I am trying to quit.

Korsgaard’s account is perhaps too rationalistic, concentrating on the practical reasoning aspect of moral agency. I prefer an account of agency that considers the role of the moral sentiments as well. And, as we have seen, the emotions, though integral to moral agency, are such by virtue of their autonomic nature; they lie largely outside the sphere of practical reasoning as such.

Identity is Forward-Looking.

It should be noted that at the core of Korsgaard’s critique of Parfit is a feature of identity and agency that tends to get overlooked. Traditionally, when philosophers have discussed the hoary old problem of personal identity, they have tended to focus on the criterion of memory, which is essentially backward-looking. Parfit may be included in this characterization. He is the acknowledged master of the potted philosophical thought-experiment, and almost all of his thought-experiments stress discontinuities of memory and experience. However, as Richard Sorabji has pointed out (and as Korsgaard’s critique implies), future plans, or forward-looking criteria, may be even more important in the constitution of identity than backward-looking criteria. Sorabji illustrates this with the example of one of the patients of the famous neurologist, A. R. Luria. This patient was shot in the head during the Second World War and as a result had lost all memory of who he was. He made it his personal project to regain his identity. He kept a diary, writing at the painstaking pace of a word or so a day, documenting his struggle to regain himself. As it turns out, this exercise was itself instrumental in giving this man an identity: “Luria comments that those of his patients who lost the ability to plan future projects disintegrated far more than those who had lost their memories. For tranquility, no doubt, this man would have had to succeed in remembering his past. But to have a firm identity, the continuing project was enough” (Sorabji 176). Korsgaard’s account has the advantage of stressing this forward-looking aspect of identity and agency.


AINSLIE, George. “Beyond Microeconomics: Conflict among Interests in a Multiple Self as a Determinant of Value,” in Jon Elster (ed.), The Multiple Self. Cambridge: Cambridge University Press, 1986.

KORSGAARD, Christine M. “Personal Identity and the Unity of Agency: A Kantian Response to Parfit,” Philosophy and Public Affairs 18 (1989), 101-132.

PARFIT, Derek. Reasons and Persons. Oxford: Clarendon Press, 1984.

RADDEN, Jennifer. Divided Minds and Successive Selves: Ethical Issues in Disorders of Identity and Personality. Cambridge, MA: MIT Press, 1996.

SORABJI, Richard. Self: Ancient and Modern Insights about Individuality, Life, and Death. Chicago: University of Chicago Press, 2006.

TAYLOR, Gabrielle. Pride, Shame, and Guilt: Emotions of Self-Assessment. Oxford: Clarendon Press, 1985.