What's the use of utility?

Philosophy and Public Affairs

Volumn 29, Issue 2, 2000, Pages 113-136

  Millgram, Elijah ^a  

^a NONE

Author keywords

[No Author keywords available]

Indexed keywords

EID: 0011075502     PISSN: 00483915     EISSN: 10884963     Source Type: Journal    
DOI: 10.1111/j.1088-4963.2000.00113.x     Document Type: Article

References (47)

1. The move from 'psychological' to 'in virtue of a class of mental states' is evidently optional. The Kantian preference for autonomy over heteronomy expresses the same or a similar thought about the direction of explanation, but lack of contradiction in the will is not a mental state on a par with desire or preference. Again, advocates of identity-based reasons (the foremost of whom happens also to be a Kantian: see Christine Korsgaard, The Sources of Normativity [Cambridge: Cambridge University Press, 1996]) take the stake you have in some things to be a matter of your practical identity, e.g., of your being a Fiat employee, a skinhead, a Croatian, or an enemy of so-and-so. Having the stake turns out, on such an account, to be a matter of one's psychology, but not of something on a par with a desire or preference.
2. Call a moral theory utility-driven if it recommends actions or policies on the basis of the difference they make to persons' utilities. Utilitarianism is perhaps the most prominent example of a utility-driven moral theory, but it is not the only one: egoistic hedonism, which recommends maximizing one's own utility, is utility-driven, as are theories concerned with the distribution of utility rather than its maximization, e.g., those requiring policies to maximize the utility of the worst-off. Theories that replace maximizing with satisficing can count as utility-driven as well. Although for expository reasons I will direct the upcoming argument against one form or another of utilitarianism, it will also cut against utility-driven moral theories that share an understanding of utility with those versions of utilitarianism.
3. While utilitarianism does have a proud history of willingness to overturn received moral views, utilitarians have never been ready to give up the idea that their moral theory was (at least indirectly) relevant to choices of the familiar kinds: Bentham and Mill were social reformers. I will return to the question of how deeply revisionist a utilitarian can afford to be in Section II.
4. J. S. Mill, "Utilitarianism," in his Collected Works, vol. 10, ed. John Robson (Toronto and London: University of Toronto Press and Routledge Kegan Paul, 1969), pp. 203-59, at p. 214.
5. Philip Brickman, Dan Coates, and Ronnie Janoff-Bulman, "Lottery Winners and Accident Victims: Is Happiness Relative?" Journal of Personality and Social Psychology 36 (1978): 917-27.
6. Because the study confined itself to interviewing accident victims within one year of their accidents, it is unclear from the study whether the hedonic effects of the accidents wear off entirely with time. For a recent overview of the field of 'hedonic psychology' (formerly called 'happiness studies'), see Daniel Kahneman, Ed Diener, and Norbert Schwarz, Well-Being: The Foundations of Hedonic Psychology (New York: Russell Sage Foundation, 1999).
7. The tendency has from time to time been mentioned in discussions of utilitarianism, as when Michael Macmillan remarks on the "vulgar saying, that we can get used to anything as eels get used to being skinned." Or again: "Most boys thoroughly enjoy eating jam tarts. But allow a boy to eat jam tarts at every meal in the day, and he will soon cease to regard them as very delightful." (Macmillan, The Promotion of Happiness [London: Swan Sonnenschein, 1890], pp. 7f; I am grateful to Christoph Fehige for bringing these passages to my attention.)
8. For instance, it might be alleged that utility-generating episodes tend to make utility come more easily in the future, by enhancing one's self-esteem, making one more of an optimist, and so on: even if utility comes in short bursts, increasing the frequency and magnitude of the bursts, and changing the ratio of pleasant to unpleasant episodes, can produce significant alteration in utility over the long term. (That objection deserves to be put more carefully than it usually is; we need to distinguish what you are - which is built up over time, and may include such things as one's self-confidence - from ephemeral at-a-time satisfaction.) This is a good place to consider the more general form of the objection, that my denial of the Presumption of Effectiveness fails to take into account downstream utilities, and that when these are factored in, the differences made to utility will generally prove proportionate to the significance of the decision. Hiring so-and-so will produce just a brief utility spike now, but once he is on the job, his decisions and actions will affect many other people, and those effects on utility account for the significance of the hiring decision. The objection is unsuccessful for at least two reasons. First, and most familiar, appeal to distant effects is not usually available to the utilitarian, because, as Kant pointed out, no one knows what they are. (See his Grounding for the Metaphysics of Morals [Indianapolis: Hackett, 1785/1981], at Ak. 418.) The distant and unknown effects generally swamp the known effects, so invoking downstream utilities will lead the utilitarian into the dead end of a moral theory that can (almost) never be applied, and the claim that the differences in utility will prove proportionate to the significance of the decision cannot be convincingly supported. Second, and more interestingly for our purposes, the appeal to downstream utilities simply pushes the problem we are considering downstream. The decisions the new hire will make are indeed more important than the utility spike he now experiences; but we will again find that their importance cannot be accounted for in terms of the negligible changes in utility that they produce.
9. Plato seems at one point to have held a view similar in some respects to the one I am now developing: that pain indicates change away from, and pleasure change toward, bodily homeostasis; the state of balance that is one's goal feels like nothing at all. See Philebus 31bff. Indicators are of course not always reliable, but I don't want to enter now into the question of how reliable this one is.
10. It follows that there are after all strategies for producing a permanently elevated hedonic tone. (See Bill Watterson, It's a Magical World [Kansas City: Andrews and McMeel, 1996], p. 40.)
11. Evidently, this can be effected by a diet of constant improvement, and situations can be constructed to provide such a diet. The 'flow'-inducing tasks described by Mihaly Csikszentmihalyi (Flow: The Psychology of Optimal Experience [New York: Harper-Collins Publishers, 1990], esp. pp. 48ff) provide a series of challenges, each successive challenge harder than the previous, but each obstacle surmountable, and when surmounted, immediately seen to be surmounted. (Think of the way video games standardly provide levels of difficulty; as soon as you have mastered level six, it's on to the slightly harder level seven.)
12. Notice that the advocate of 'flow' has an answer - but an unsatisfying one - to the charge that, while he can produce largish effects on utility, these do not allow us to make sense of the practical questions we already have. 'Flow'-generating activities are completely absorbing, and one's unanswered practical questions may just drop away of their own accord. The compulsive gambler or Nintendo addict, having lost sight of everything beyond the next game, may occupy a perspective from which the Contoured Presumption of Effectiveness is perfectly in order. In my "Pleasure in Practical Reasoning," The Monist 76 (1993): 394-415,
13. and in Practical Induction (Cambridge: Harvard University Press, 1997), sec.6.6,1 tried to explain away the diminishing returns of familiar pleasures. I now think that the examples of enduring pleasures that I gave can be accounted for in terms of the increasing levels of, e.g., complexity, nuance, and challenge that appear as projects and tastes develop, that is, by the presence of a regular diet of improvement. Without improvement, pleasures generally diminish, and this suggests that I was mistaken in treating pleasure as tantamount to a judgment of desirability, as opposed to something like a judgment of improvement.
14. One possibility worth considering is that the indicator is responsive not just to change in well-being, but also to rate of change: that the connection between utility and welfare is the connection between the first derivative of a function, and the value of that function. See Christopher Hsee, Robert Abelson, and Peter Salovey, "The Relative Weighting of Position and Velocity in Satisfaction," Psychological Science 2 (1991): 263-66;
15. Hsee and Abelson, "Velocity Relation: Satisfaction as a Function of the First Derivative of Outcome Over Time," Journal of Personality and Social Psychology 60 (1991): 341-47.
16. For a standard, if dated, introduction, see Patrick Winston, Artificial Intelligence (Menlo Park: Addison-Wesley, 1977), pp. 93-98.
17. Whether hill-climbing is an appropriate technique depends, among other other things, on whether the space of options has the requisite nearness or adjacency relations. I suggested thinking of the distance of the options from one another as how difficult it is to get there from here. Sidling around to determine which of several easily reachable points in a landscape is the highest is a reasonable strategy if it's easy to undo the last step, that is, if the distance from A to B is the same as the distance from B to A. But of course that's often not true: it can be very easy to take steps that it's very hard to undo. In "Moral Values and Secondary Qualities," American Philosophical Quarterly 36 (1999): 253-55, I argued that the fact that many moral reactions extinguish over time is a problem for secondary-quality accounts of moral values; the moral significance of a situation, after all, is not supposed to evaporate simply because we get used to it.
18. (Gilbert Harman, "Critical Review: Richard B. Brandt, A Theory of the Good and the Right," Philosophical Studies 42 [1982]: 119-39, at p. 128, makes a related point against Brandt's use of 'cognitive psychotherapy' as a test of the rationality of desire; I am grateful to Christoph Fehige for bringing it to my attention.) But the point evidently has to be hedged; we can treat a rapidly extinguished response as constitutive of a stable secondary quality provided we can see it as an indicator of change in the secondary quality. I think that a secondary-quality treatment of objective welfare along these lines should not be too hastily ruled out, but I doubt that many of the so-called moral values will be amenable to this kind of reconstruction.
19. Both the feelings that are the converse of elation and the like, and a familiar category of bodily sensations, are called painful. While it is not just a coincidence that painful feelings are so-called, the sensations and the feelings nonetheless differ, and I am unsure how far the account I am developing holds good for the bodily sensations; chronic back pain is (as a reviewer helpfully reminded me) indeed chronic.
20. The indicator account of phenomenal utility has no problem making room for the instrumental importance of pleasure and displeasure. The sensitivity of one's hedonic tone to circumstances may be important, in the way that other well-functioning indicators are important: because how fast you're going matters, that the speedometer is working also matters. Notice that this can produce situations in which hedonic valence fails to match the direction of change in one's welfare. An example due to Allen Coates: in discovering that one's marriage had been a sham, one feels worse, because one's hedonic indicator is now reflecting how bad things really are, but one is arguably doing better, because recoupling one's various indicators to the world is the first step toward improving one's situation. The account can also allow that even if we are primarily interested in changes in our welfare because changes affect the absolute level of welfare, the changes themselves might come to be valuable. Speedometers indicate not distance traveled, but speed, the first derivative of distance with respect to time. Our primary or first interest in speed, and so in the speedometer, is in distance traveled: "Are we there yet?" "When will we be there?" But we can come to have secondary interests in speed proper: for the flow of traffic to be coordinated, vehicles must move at approximately the same speed; driving too fast will get you a ticket; and some people just like driving fast.
21. There are exceptions, e.g., Leonard Katz, "Hedonism as Metaphysics of Mind and Value" (PhD Thesis, Princeton University, 1986).
22. Friedrich Nietzsche, Beyond Good and Evil, trans. Walter Kaufmann (New York: Vintage Books/Random House, 1886/1966), sec. 18.
23. If I am right, however, in thinking that this view best captures the spirit of the subjectivist approach to utility, it is likely to be relevant to the assessment of other, actually articulated satisfaction-based conceptions of utility.
24. Someone who holds such a view may hold that satisfaction of a preference or desire is a contribution to an individual's utility, whether or not the individual knows the preference or desire to have been satisfied. As we will shortly see, however, knowledge of satisfaction is by no means irrelevant to the way utility of this kind functions.
25. Note, first, that by "live" I do not mean 'phenomenally vivid.' Second, there is a widely used technical notion of utility on which the utilities of outcomes are constructed from the agent's preferences, and I need to emphasize that this is not the type of utility we are now starting to examine. For one thing, expected utility theory requires preferences that are, from the perspective of the view we have on the table, extravagantly numerous: over all outcomes, taken pairwise, and also over probability mixtures of outcomes. For the canonical exposition of EU theory, see R. Duncan Luce and Howard Raiffa, Games and Decisions (New York: John Wiley and Sons, 1957), chap. 2. Notice that on the live-preference-satisfaction view we are considering, there are two ways in which one may have a stake in the means to further things about which one cares. They may be of import to one, although one has not come to have live desires for them, as means or instruments; alternatively, one may have invested psychological resources in the now-live instrumentally derived desires.
26. I have found that a very common response to this claim is incredulity, cited together with one instance or another of an event that either has made all the difference (moving to California, land of good weather, has come up more than once), or that would make all the difference if it happened. If you have reactions of this kind, I am going to ask something rather difficult of you: to discount them. The reason is that it turns out that people are very bad both at forecasting affect and at remembering how well they felt. To use the example I was given, by someone who was convinced that moving to sunny California had made her life much happier than it had been: a study by Kahneman and Schkade "found no difference in self-reported well-being between students at California and midwestern universities, despite large differences in satisfaction with their respective climates . . . [but] students predicted large differences across regions in both overall well-being and in satisfaction with the climate" (George Lowenstein and David Schkade, "Wouldn't It Be Nice? Predicting Future Feelings," in Well-Being, ed. Kahneman, Diener and Schwarz, pp. 85-105).
27. (For a disconcerting example of unreliable hedonic memory, see D. Kahneman, B. Fredrickson, C. M. Schreiber, and D. Redelmeir, "When More Pain Is Preferred to Less: Adding a Better End," Psychological Science 4 [1993]: 401-5.
28. For current work on affective forecasting, see Daniel Gilbert, Elizabeth Pinel, Timothy Wilson, Stephen Blumberg, and Thalia Wheatley, "Immune Neglect: A Source of Durability Bias in Affective Forecasting," Journal of Personality and Social Psychology 75 [1998]: 617-38,
29. and Daniel Gilbert and Timothy Wilson, "Miswanting: Some Problems in the Forecasting of Future Affective States," in Feeling and Thinking: The Role of Affect in Social Cognition, ed. Joseph Forgas [Cambridge: Cambridge University Press, 2000], pp. 178-97.) Certainty in how one would feel, or in how one did feel, is, it turns out, quite often misplaced. I am of course not the first to have noticed that desires and preferences change in response to changes in the availability of their objects.
30. See, e.g., Jon Elster, Sour Grapes (Cambridge: Cambridge University Press, 1983).
31. One's desires and preferences may be reenlivened when the prospect of losing their objects is made vivid, and one once again comes to care; but this does not mean that one cared all along. One may, and this is very important, have a subjective stake in something counterfactually: although I have taken my house completely for granted for a long while, if losing it were to become a live option, my desire to keep the house would be enlivened on the spot. (I am grateful to David Lewis for this example.)
32. There is a related reason for the failure of still another form of utilitarianism, one that demands the satisfaction not of actual but of informed desires or preferences. Informed-preference utilitarianism is motivated by the idea that what one now wants may turn out to be disappointing when one finally gets it, and so actual preferences or desires must be corrected before they can be used as a guide to action; the standard deployed is the preferences and desires the agent would have in an appropriately idealized situation. Various idealizations have been suggested at one time or another: being better informed, calm, and collected, or a beneficiary of 'cognitive psychotherapy.' But since what we really want to know is what the object of the desire will be like, to the agent, when it is obtained, the standard idealizations are really attempts to introduce information you're generally not in a position to have until you've got the object of your desire in hand. Often, the only way to know what things are like is to have tried them out, and so the right appeal has to be to the preferences and desires of a counterfactual version of oneself who has had them satisfied. The right choice is the choice that would be made by a version of you that has already made the choice both ways, and seen how each comes out; his word is the final word. Now what will that final word be? What would an agent feel about the satisfaction of some candidate desire, given that he has already had the desire satisfied? We are seeing it to be a deep fact about the mental processes which keep our supply of motivating states replenished that what we have attained typically ceases to be desired by us: when a desire is satisfied, our attention turns to a different, perhaps newly formed desire, because this desire is able to guide action and the satisfied desire is not. So when the candidate desire is satisfied, interest will fade for that very reason, and we will come to prefer something else - something that we perhaps do not yet want and that we as yet have no reason to pursue. This means that appeal to counterfactually informed desires will give us systematically incorrect results: it will tend to tell us not to pursue what now matters to us, in favor of what might someday (but does not - and should not - yet) matter to us. And this means that a utilitarianism of informed desires and preferences, the point of which is to give us what we really (rather than what we actually) want, will, for the most part, give us something we don't really want at all. That said, there are many variants of informed-desire theory, and not all of them are directly addressed by this point.
33. A sociobiologist or evolutionary psychologist might add, and if we are to have a good chance of leaving our genes to the next generation. When asking what desires, preferences, and so on are for it is very easy to come up with an answer like: increasing reproductive success. But this is evidently too fast an answer. The point of our earlier discussion of phenomenal utility was that we should be careful to avoid mistaking the indicator for the item of interest. A similar caution is in place when we are examining changing gene distributions in a population: is this the item of interest, or is it rather a device that regulates, more or less effectively, the items of interest?
34. There is an analogous point to be made here about democratic politics. I have sometimes heard it said that the point of a democratic government is to satisfy the preferences of the citizenry, the implication being that in an ideally functioning democratic state, the preferences of the citizenry would be almost always satisfied. But this mistakes the function of the voting populace's preferences in a democratic regime. In a healthy democratic polity, the government gets kicked out fairly regularly, and, during its stint as the opposition, is forced to come up with new agendas, new leaders, and new ideals. For this to be possible, voters must develop unsatisfied preferences when the party in power is satisfying the preferences they already had. In a well-functioning democracy, a largish fraction of the voters are dissatisfied with their government, and it is practically impossible to raise the satisfaction level over the long term.
35. The transition was not without its lurches; for a partial history, see Robert Cooter and Peter Rappoport, "Were the Ordinalists Wrong About Welfare Economics?" Journal of Economic Literature 22 (1984): 507-30.
36. Ford, Independence Day (New York: Vintage/Random House, 1995), pp. 38f. Note the contrast between this very realistic and rich specification of a desire, and the oh-so-thin typical philosopher's example!
37. Of course, there may be other ways people adjust their preferences, not involving thresholds; I don't want to insist that this is always what is going on.
38. Satisficing was originally introduced as a second-order maximizing strategy, one that took computational and information costs into account (Herbert Simon, Models of Man [New York: John Wiley and Sons, 1957], chaps. 14, 15).
39. But there is a deeper reason for adopting it. It has been pointed out (by, e.g., Chris Landesman, "When to Terminate a Charitable Trust?" Analysis 55 [1995]: 12-13,
40. and Christoph Fehige, "The Limit Assumption in Deontic (and Prohairetic) Logic," in Analyomen 1, ed. Georg Meggle and Ulla Wessels [Berlin: de Gruyter, 1994]), that maximizing strategies make no sense when one's choice sets do not have maximal elements. Now, if you think about it for a moment, that is, for all practical purposes, almost always one's situation; a little looking around will, surprisingly often, turn up an alternative better than those already under consideration. (I am grateful to Jim Fearon - who does not at all share my view here - for conversation on this topic.)
41. For an overview of an older related literature, see William Starbuck, "Level of Aspiration," Psychological Review 70 (1963): 51-60.
42. The connection we have just made may change the way we see some of the available accounts of pleasure. Brandt's view, for instance, "that for an experience to be pleasant is for it to make a person want its continuation" (Richard Brandt, A Theory of the Good and the Right [Oxford: Clarendon Press, 1979], p. 38) is evidently - although not simply wrong - badly off-base. Part of what pleasure does is to tell you to keep doing what you're doing; but another primary role of pleasure is to extinguish, rather than prolong, desire.
43. See, e.g., Rudolph Carnap, The Logical Syntax of Language (London: Kegan Paul, Trench, Trubner and Co., 1937), sec. 74.
44. Instrumentalism can also be expressed in the material mode, as the 'fact-value distinction,' or one version of it, and in the technical mode, as one very widespread way of endorsing and interpreting the apparatus of decision theory. Let me qualify my analogy to the logical positivists' 'modes.' The logical positivists rendered their claim of equivalence between items expressed in different modes as the claim that it is possible to translate from one mode to another. I am not making this claim, but I do not think it necessary. Jack loves Jill: he may send her flowers, or he may send her chocolates, or he may send her love letters. Flowers cannot be translated into chocolates, or chocolates into love letters, but we understand what is being claimed when we say that these are different ways of expressing the same thing.
45. If I am right in thinking that belief-desire psychology, instrumentalism stated in terms of goals, the (or a) fact-value distinction, and a very common interpretation of expected utility theory are all just different ways of expressing the same view, related to each other as were the positivists' modes, then there is a very important consequence for debates about instrumentalism: using any one of these as a premise from which to prove the other is an argument on a par with stating one's claim in English, repeating it in Dutch, and finally reiterating it in German. That is to say, putative arguments of this form are not in fact arguments at all. Instances of the misguided would-be argument pattern include Michael Smith, "The Humean Theory of Motivation," Mind 96 (1987): 36-61,
46. and David Lewis, "Desire as Belief," Mind 97 (1988): 323-32.
47. As it happens, this fragmentary view of practical reasoning is evidently not instrumentalist, so if a mental ontology that includes desires is in fact an expression of an instrumental theory of practical reasoning, then, although I have been conducting the discussion in the accepted idiom of desires, this should be regarded as a temporary concession to current usage, ultimately to be abandoned.