Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!elroy.jpl.nasa.gov!ncar!gatech!udel!princeton!pucc!PSYCH@TCSVM From: harnad@phoenix.Princeton.EDU (Stevan Harnad) Newsgroups: sci.psychology.digest Subject: PSYCOLOQUY V2 #1 (Baron on Optimal Utilities/Becker 250 lines) Message-ID: <9101040108.AA10494@reason.Princeton.EDU> Date: 3 Jan 91 23:41:17 GMT Sender: VMNNPOST@pucc.Princeton.EDU (Listserv to Netnews Gateway) Organization: Listserv to Netnews Gateway at pucc.Princeton.EDU Lines: 245 Approved: PSYCH@TCSVM PSYCOLOQUY Thu, 3 Jan 91 Volume 2 : Issue 1 Jonathan Baron: Comments on Gordon Becker's Optimal Utilities paper [Editor's Note: This commentary on Becker's Optimal Utilities (Psycoloquy 1 (16) 1990) has been refereed by a member of Psycoloquy's Editorial Board and has been accepted for "skywriting" discussion. All discussion elicited on this topic will appear under the heading: "Optimal Utilities/Becker." Comments as well as discussion papers on other topics are invited. All contributions will be refereed.] ---------------------------------------------------------------------- From: baron@cattell.psych.upenn.edu (Jonathan Baron) Subject: Comments on Optimal Utilities paper by Gordon Becker Cc: baron@cattell.psych.upenn.edu, becker@zeus.unomaha.edu Status: R The rationality of the choice of goals: Comment on Becker(1) Jonathan Baron Department of Psychology, University of Pennsylvania Becker argues that tastes, or, as I shall call them, goals, are changeable and that utilities can be maximized by changing tastes so that utility is maximal for all possible outcomes. Changing tastes has costs in utility, however, so equal utilities should be seen as a goal toward which we strive rather than something we can easily achieve. Buddhists strive for such equanimity, and, perhaps, achieve it to some extent. Let me first dispense with a minor problem with Becker's formulation. He suggests that utility can be maximized by setting all utilities equal to the utility of the best outcome. But that would mean setting the utility of the worst outcome equal to that of the best outcome, since the worse outcome is an outcome among others. All utilities would then be equal to the utility of the worst outcome, so the outcome that occurred, whatever it was, would be the worst possible outcome as well as the best possible outcome. The argument for setting all utilities equal appears weaker in view of this move. But Becker's argument can be reformulated to avoid this problem. He could argue that we should increase our desire for what we have (or are likely to get) and decrease our desire for what we do not have (or are not likely to get). The prescription that we should change our tastes to conform to what we have has been discussed by Elster (1983) in a book aptly titled "Sour grapes." Elster was inclined against such changes of preferences, on the grounds that they discouraged the sorts of striving that might lead to large improvements, although Elster was also aware of the advantages of such "adaptive preference formation." Elster's and Becker's arguments raise a more general question, which is how we can conceptualize changes of desires and incorporate them within utility theory as a normative model (if we can do so at all). I have suggested (Baron, 1988, pp. 430-432) that the choice of goals (or desires) is itself a decision problem. Or, more precisely, we must decide what actions to take to TRY to influence our goals. A normative theory of decisions about goals (including strengthening, weakening, adding, or deleting) can be based on the same criterion as that applied to other decisions, the maximization of utility, or, in other words, the greatest achievement of (other) goals. Some of our goals concern our goals themselves. We want goals that others approve of, or goals that will give us certain good feelings when we try to achieve them or succeed in achieving them. These "goals for goals" generate utilities when we make the choice about goals. Decision about goals also affect directly the achievement of other goals that we already have. This is another way in which decisions about goals have utility. Finally, choices about goals affect the achievement of goals that we will have in the future (whether these goals will arise inevitably or as a result of our present decision about goals). One consequence of many choices, aside from those designed purely to change our goals, is that they affect our goals - increasing the strength of some and decreasing the strength of others - and we should take this into account. In some decisions, we must simultaneously consider the modification of goals and their satisfaction. For example, I have considered running for congress, expecting not to win but to "educate" my fellow citizens about certain issues. If I did this, my desire to win would probably increase as the race went on. In making my decision whether to run, I must consider simultaneously the creation of this new goal and the (low) probability of its achievement. It would be a frustrating experience, unless I could, through rigorous self-control, prevent myself from wanting to win. We are tempted to think that decisions of this sort involve a certain incommensurability. For example, the decision to have a child involves a choice of two different paths of life - a choice of becoming two different people. One person has the goal of caring for a child and the other has stronger goals concerning the companionship of friends, entertainment, or achievement. One commonly mentioned consequence of having children is that people CARE less about aspects of their lives that they have to give up in order to find the time to be a good parent. It is not just a matter of going against one's goals. Rather, after some period of time, the goals weaken, and the failure to satisfy them is no longer experienced as a great frustration. How should we compare two paths of this sort? Of course, one source of difficulty is the enormous uncertainty we have about how much each goal will be satisfied and how strongly we will experience it, but we must put this uncertainty aside. It is not the issue. (It is, in any case, not peculiar to this type of decision.) We must, as in other decisions, make our decision on the basis of probabilities as we know them. What is at issue is the comparison of different goals with different levels of achievement of each goal. We may simplify the decision as follows: Option 1: Have child Goal of child's welfare expected strength: 90 expected level of satisfaction: 70 Goal of social life expected strength: 40 expected level of satisfaction: 20 Option 2: Do not have child Goal of child's welfare expected strength: 0 expected level of satisfaction: 0 Goal of social life expected strength: 80 expected level of satisfaction: 60 How should we compare these two options? Is it better to have the child-welfare goal partly satisfied, as in the first option, or not have it at all, as in the second? What if the expected level of satisfaction were 40 or 10 instead of 70? Of course, these numbers are just stand-ins for real consequences, and the scale is arbitrary. One way to make these comparison is by consulting the goals we have that remain constant across the two options. In doing this, we apply the idea of utility as goal achievement to goal-plus-satisfaction outcomes such as those just described. We may, for example, decide that it is better to have the goal of child welfare, partly satisfied, because it fits with our other goals of concern about other people and the future and having a full range of human experiences. But we may decide that the expected level of achievement of the goal - given the conditions in which the child would have to grow up, say - might be "negative," that is, worse than the alternative of not having the goal and not having it satisfied. Such a conclusion might result from goals not to make people suffer from what we see as a bleak future. In reality, such judgments are difficult to make, even if we are certain of the consequences we are judging. But they are not IMPOSSIBLE to make. It makes perfect sense to say, "I wish I didn't care so much about what happens to Susan, given what is happening to her." Note that this method of evaluation does not allow us to reason from any particular first principles (unless such principles can be independently justified). We cannot compare two sets of goals without some some core set of common goals. In this respect, the evaluation of goals is similar to the evaluation of beliefs in the Bayesian theory. We can evaluate the probability of each belief, given the probabilities assigned to all other relevant beliefs, but we cannot compare systems of belief as a whole. Could we do more than this for goals? Could we, for example, include each goal as part of its own evaluation? Suppose I am faced with a choice that involves the development of new goals. For example, suppose I consider taking up watercolors as a hobby. In evaluating each such activity, can I also take into account the goals that result from the activity itself? I do not mean such general things as "being successful at things that I do," for this is a goal I already have, which I can use as a standard to evaluate new goals. I mean things inherent to the activity, such as adhering to certain standards of watercolor design, standards I do not even know about right now (except insofar as they derive from more general aesthetic standards). I see no way of making such evaluations. To make any comparative evaluation, we need common criteria (that is, common goals, since a goal is a criterion of evaluation). If I am right about this, our ability to make comparisons of two courses of life is dependent on the existence of common goals. In conceiving of goals and their rational adoption, it might help to think of each goals as a legislator in a governing oligarchy (of "multiple selves"). Each legislator has a FIXED agenda, a set of criteria (goals) for evaluating every proposal put before the group. Admission of new members is based on the same agendas. Voting is not used; instead, the honest appraisals of each member are added up. That is, the group admits a new member when the expected behavior of the new member furthers the agendas of the members more than does not admitting the new member (or more than does admitting some alternative possible member). But, importantly, the new member is not simply a means to further the agendas of the current members, although that is why they admit her. The new member brings an agenda of her own, thereby changing somewhat the overall behavior of the group. In this way, the rational adoption of goals is instrumental, but its effect is not solely instrumental. New goals are truly added. If we think of the strengthening or weakening of goals as anlogous to addition or subtraction of new (identical) goals, then we can easily extend this argument to the sorts of problems that Becker considers. References Baron, J. (1988). Thinking and deciding. Cambridge University Press. Elster, J. (1983). Sour grapes: Studies in the subversion of rationality. Cambridge University Press. Note 1. Parts of this comment are excerpted from a manuscript in preparation. ------------------------------ PSYCOLOQUY is sponsored by the Science Directorate of the American Psychological Association (202) 955-7653 Co-Editors: (scientific discussion) (professional/clinical discussion) Stevan Harnad Perry London, Dean, Cary Cherniss (Assoc Ed.) Psychology Department Graduate School of Applied Graduate School of Applied Princeton University and Professional Psychology and Professional Psychology Rutgers University Rutgers University Assistant Editors: Malcolm Bauer John Pizutelli Psychology Department Psychology Department Princeton University Rutgers University End of PSYCOLOQUY Digest ******************************