Baron, J. (1996). Norm-endorsement utilitarianism and the nature of utility. Economics and Philosophy, 12, 165-182.

Norm-endorsement utilitarianism and the nature of utility

Jonathan Baron1
University of Pennsylvania

Norm-endorsement utilitarianism and the nature of utility

In this article, I shall suggest an approach to the justification of normative moral principles which leads, I think, to utilitarianism. The approach is based on asking what moral norms we would each endorse if we had no prior moral commitments. I argue that we would endorse norms that lead to the satisfaction of all our nonmoral values or goals. The same approach leads to a view of utility as consisting of those goals that we would want satisfied. In the second half of the article, I examine the implication of this view for several issues about the nature of utility, such as the use of past and future goals. The argument for utilitarianism is not completed here. The rest of it requires a defense of expected-utility theory, of interpersonal comparison, and of equal consideration (see Baron, 1993; Broome, 1991).
I reject moral intuition as justification for moral claims. Moral intuition is too easily understood as arising from the application of heuristics that are usually good but often misapplied (Baron, 1994a). For example, the distinction between acts and omissions is morally relevant in a variety of situations, but some people apply it in ways that others see as erroneous: for example, some people are reluctant to vaccinate children against diseases when the vaccine can cause harm, even though it reduces substantially the overall probability of harm (Ritov & Baron, 1990; Baron, 1994a; Asch et al., 1994). Philosophers are not necessarily immune from errors based on misapplication of principles. In general, such empirical results give us another interpretation of our intuitions, against the view that intuitions arise because of some connection with truth.
Instead of relying on intuitions, I try here to justify normative moral principles in terms of their purpose, which is ultimately to regulate each other's conduct in accord with our goals. Such principles are expressed through our endorsement of norms. By asking what norms we would endorse to achieve this purpose, we can evaluate alternative normative proposals. Such a method is a kind of analysis. It carves off a domain of human activity and labels various parts of it (e.g., thinking, decision making, goals, options, outcomes). The argument is therefore of the form: if we think about things this way, and if we have this set of basic goals, then it follows that the following principle will help us achieve these goals.
The specific form of utilitarianism that I shall defend is one that concerns itself with maximizing the achievement of goals or the satisfaction of values as determined by the individuals who hold those values. The primary purpose of this kind of utilitarian theory is to guide decision making rather than to justify praise or blame (except that acts of praising and blaming are also decisions with consequences). Moreover, I distinguish between normative principles - which are concerned with ultimate justification - and prescriptive principles - which are practical guides to action (Baron, 1994b). This article is about normative issues.
By goal or value I mean a standard for evaluation of outcomes, as assumed in decision analysis (e.g., Keeney, 1992). In theory, and perhaps in practice, goals are best measured by asking people to make judgments of states of affairs, not by observing choices. Goals are by definition the standards that we have for evaluating how things turn out. Thus, conflict between choices and goals is prima facie evidence that the choices are in error. (The term "preference" can refer to choices or values.) The idea of utility as goal achievement allows us to count goals that some theories do not count, such as past goals. But it is does not allow us to count such goals as interpersonal equity (as done by Broome, 1991, and Hammond, 1988) unless these can be reduced to the goals of individuals.
I make no important distinction between a goal and a set of goals. Almost any criterion can be subdivided. For example, the criterion "spells correctly" can be subdivided into different kinds of spelling patterns or even different words. Likewise, goals can be combined. So, if a state can be evaluated by each of several goals, it can also be evaluated by all the goals in combination. We may therefore speak of the achievement of a set of goals taken together. This assumption is necessary if the idea of goal achievement is to yield decisions.

Norm endorsement

Morality is often taken to concern what we should do, in some sense of "should." But I think we can make progress by narrowing the topic somewhat. Instead of looking at what we should do, let us look at the social function of morality only. Specifically, morality involves telling each other what we should do, or otherwise trying to influence each other's decision making. Moreover, the telling goes beyond specific cases. When we try to influence people morally, we try to influence some standard they hold for how they should behave in a class of cases (possibly a small class, but not restricted to a single instance). We thus try to influence the principles they follow, not just their behavior in a specific situation.
I shall call this telling function "norm endorsement," using Gibbard's (1990) term, without necessarily accepting everything he says about it. I emphasize the social function of endorsement - which Gibbard referred to as "normative discussion" - but I take no position about Gibbard's claims concerning the centrality of moral emotions such as guilt and anger. What matters here is whether behavior can be influenced by advice or by the promulgation of standards. Morality as I discuss it is useless if it cannot be used as a way of giving advice or of inculcating standards that will be followed to some extent. It does not matter here how this influence works, or whether the rationality of advice giving can be reduced to questions about what emotions it makes sense to feel.
Norms, in this view, are standards of choice or behavior that people try to induce each other to follow. When we give advice, exhort others to behave in certain ways, or advocate certain principles, we endorse certain norms. These norms can either be prescriptive guides or normative standards. Asking what norms we have most reason to endorse (for either purpose) is a way to think about what normative or prescriptive principles are best.
We can express morality, as a way of influencing others, in several ways: teaching principles directly by explaining them or by expressing moral emotions; setting examples of how to behave; gossiping (and thereby expressing principles by presupposing them when we criticize others - see Sabini & Silver, 1982); rewarding behavior that follows the principles and punishing behavior that violates them; supporting or violating institutional (family, governmental) laws or rules (depending on what we think of their morality); supporting or opposing such laws or rules, making or repealing them; or justifying them in hindsight by referring to norms, which are then supported by the appeal. The concept of morality can be seen as arising as an abstraction from all of these situations (as suggested by Singer, 1982). We can think of the purpose of moral standards as advice giving, with a view to affecting decisions, although the advice need not be verbal.
Suppose that there is such a thing as a "morally best set of norms" that is not what we have the most reason to endorse for each other to follow. People's reasons for endorsing norms lead them to endorse some other set, not this set. There are then no reasons to which we can appeal in order to persuade each other to endorse this "best" set. It is out of the question, unless we are irrational. If "ought" implies "rationally can," then this set fails.
Moral exhortation does not include all attempts to impart standards of conduct to others. "Please don't wear that perfume around me," for example, is an expression of personal preference, not an endorsement of a moral norm. Moral principles are expressed by groups and are taken to be the same for all members of the group. This property makes them "universal" in Hare's (1963, 1981) sense. They are impartial with respect to individuals in the group. (I leave aside the question of why we should consider the largest possible group.) Of course, an act of moral exhortation could serve other functions, such as advancing the self-interest of the speaker. Exhortation, then, is not so much a type of act as a purpose that acts could serve among other purposes.
So the proposal is to try to answer the question of what standards we should endorse for others and ourselves to follow, if any. A simple answer is that we should endorse the moral principles we hold. If we are committed to these principles, that is surely what we would do. This answer is not acceptable here, because we are trying to use the function of morality to answer the question of what principles we should endorse. To do this, we must put aside our previous moral commitments. More generally, we must put aside any principles that tell us what to endorse, whether we take these principles to be moral or not. Such principles might include aesthetic or personal ideals, such as piety, fashionable style, or personal development, that would not count as moral in a narrow sense but serve the same social function as moral expressions. We must put these aside because the question is what principles we should endorse. If we do not put these principles aside, then they will affect the conclusion. We are trying to come up with a set of principles from scratch.
Having put aside prior principles for endorsement, we can then ask whether we have any reasons to make any endorsements, what those reasons are, and what endorsements they lead us to make. Moreover, when you think about this, you must put aside not only your own principles but also everyone else's. Otherwise, circularity will still result, but with your principles out of the loop.
What is left? What premises can I reason from? What reasons do I have to endorse anything for others? First, I have self-interested values for others' behavior, things I like other people to do, such as being nice to me, or perhaps realizing themselves in certain ways (being pious or stylish or interesting, depending on who "I" am).
Second, each of us has altruistic values, values that depend on the personal values of others. If you like Berlioz, I can satisfy an altruistic value by giving you a Berlioz recording even if I hate Berlioz myself. Altruistic values need not concern endorsement, so, insofar as they do not, we may include them, even though they are moral in the sense of being morally good goals to have (goals we would endorse). That is, it is possible to have altruistic values without also endorsing these values as principles. I shall argue that such altruistic values are more important than selfish ones as reasons for endorsing norms.
Social morality can thus be viewed as a set of standards that we try to get each other to follow in making decisions. In asking what are the best moral principles, we might well put ourselves in the position of a group that had no principles for advice giving about decision making and then ask what principles would best serve all the remaining values of the members of the group. These remaining values, both altruistic and self-interested, would serve as the reasons for adopting moral principles. In the end, all of these values come from self-interested values, because the specific altruistic values of one person depend on the self-interested values of another. The most general moral standards are those that apply to the largest group, including people as yet unborn. Such standards do not need to be changed as a function of time and place. If we want this sort of general system, we should not base our morality on values that are peculiar to a particular group of people.
This exercise leads to a form of consequentialism. If the reasons we have for endorsing norms come from self-interested and altruistic values, then the norms we would endorse must be those that satisfy those values. For example, we would endorse a norm that says to satisfy everyone's values as best we can. Even if such a norm did not tell us how to weigh each person's values, it would not be empty. It would, for example, tell us not to endorse norms in favor of nondeterrent punishment, since this goes against the values of some but in favor of the values of none. (We have put aside values arising from prior moral beliefs, such as the belief in retribution.)
This is a skeptical view of moral principles. It challenges others to show why anything other than nonmoral goals is relevant to the adoption of these principles. Someone who values freedom, even when it goes strongly against other goals, would have to explain why we should endorse freedom if we did not already value it as something to endorse. Once we figure out the principles that best satisfy our values, then the admission of other principles will either be empty or will subvert the achievement of our goals (Kupperman, 1983).
The following exercise in imagination, while not required for the points I have made, may make them more vivid. Imagine we live in a society without any morality. Obviously, it is relatively new, or else it would have done itself in. Praise, blame, punishment, and gossip are unknown. Then someone gets the idea of having a moral system. The idea is that certain things will be counted as right and others as wrong. We will teach our children to do the right things and abjure the wrong things. We will say nice things about people who do the right things and nasty things about people who do the wrong things. We will try to stop people from doing the wrong things by "punishing" them publicly, so that others see what will happen to them if they do these things. And so on. We agree to do this. Now all we need to do is to draw up the lists for what we count as right and wrong. So each of us has the chance to say that X is right or Y is wrong. But in this imaginary world, we have no prior views. If I say "X is wrong" because I have always thought it was wrong without knowing why, you have very little reason to pay attention to me unless you already agree. I have given you no reasons. I have just asserted my view. But in the imaginary world this cannot happen because we have no prior views. We have just gotten the idea of having a morality. Thus, I must give you a reason, and I must even have one myself, or else there is no reason for me to say anything. One reason that we all have is that we care about each other. We want good outcomes for others, and we dislike bad outcomes. If I say that "infliction of suffering is wrong" out of this kind of motivation, you have reason to agree with me. A moral rule against doing harm is likely to decrease your suffering and everyone else's, and suffering is something you too would like to reduce. We can thus agree on moral rules because we already have certain desires for each others' good. The rules that we endorse and come to accept in this way will be those that, in general, bring about good consequences and prevent bad consequences. And our judgment of consequences does not require any preexisting morality.
The idea of asking what we should endorse, rather than what we should do, may seem strange. But it does capture the social aspect of our moral life, the idea that morality is public.

The motivation for making and heeding norm endorsement

People have both self-interested and altruistic reasons to endorse norms. If you convince others to do good for others generally, for example, then they will do good for you along with everyone else. However, your own altruism increases the value of norm endorsement. Suppose that you weigh the good of each other person some proportion, A, of your own good. And suppose that, by endorsing to an audience of one person the norm of doing good, you can make that person do an extra amount of good G for each of N other people, as well as for you. Then your selfish benefit for the endorsement is G, but, if we add in the altruistic benefit, then the total benefit is G + AGN. Clearly, if A is large relative to 1/N, then the second term will make norm endorsement much more valuable. Even if A is small (but positive), the second term contributes to your motivation to endorse norms.
Rational self-interest cannot motivate people to follow purely altruistic (self-sacrificing) norms that others endorse. People can be "tricked" into thinking that following a norm is in their self-interest, and both biological and cultural evolution may well have prepared us to be tricked in this way. In part (as Gibbard, 1990, suggests), we learn to follow the endorsements of others because they often involve coordination rather than self-sacrifice. It is in everyone's self-interest to heed an endorsement for the sake of coordination, i.e., when it is in each person's interest to do what others do. If someone says, "In this country, we drive on the left," then it is in everyone's interest to follow that (in the absence of contrary information).
Rational altruistic people, however, can be motivated to follow norms that others rationally endorse out of altruism. This can happen in three ways. First, the endorsement may remind people to adhere to their own previously formed altruistic goals. Altruistic goals often compete with desires that are seen as temptations, distractions from following one's better judgment.
Second, endorsements of more specific goals can provide information about the connection between pursuit of these goals as means and deeper altruistic values already present, e.g., "If you care about helping poor people, support candidate X." Thus - assuming adequate explanation - a new goal of supporting X is created as a subgoal or means of achieving the deeper altruistic goal of helping the poor.
Third, rational altruism can motivate heeding endorsements when a critical mass of cooperators is needed to provide some benefit to all, in the way in which public officials can prop up the value of a currency by saying that it is undervalued, which leads investors to buy it thinking that other investors will buy it too. If the benefit of cooperation were not provided unless some minimal number cooperated, then endorsement would encourage people to cooperate by making them believe that others would cooperate.
Of course, endorsements made out of altruism can also work for irrational reasons. People may have a biological propensity to be influenced by each other, that is, to be "docile" (Simon, 1990). Sustainable cultures may encourage this propensity in a variety of ways.
In sum, rational self-interest is limited in its ability to maintain a system based on norm endorsement. People can endorse norms out of self-interest, but they cannot expect others to heed these endorsements out of rational self-interest alone. The rational core of a system based on norm endorsement must be driven by altruism. This is why self-interested endorsements must be disguised as altruistic: "I'm telling you this for your own good," or "I'm telling you this because I know you care about Aunt Matilda." Irrational forces may help to maintain the system, but they are ultimately unstable without the rational core. People who see moral endorsement as mere expression of the interests of those who do it - like some of the radicals of the 1960s - can come to distrust morality altogether.

Conflicting principles and self-interest

Reliance on altruism helps to solve another problem, the conflict of principles. If people choose which principle to endorse on the basis of self-interest, people would endorse conflicting principles. This might not be a decisive reason to reject the present approach. We already have a world in which people endorse conflicting principles. If everyone accepted the present approach, the number of such principles would decrease, since many principles would be ruled out. Still, more reliance on altruism may allow us to do better.
Consider, for example, three sorts of rules that have been proposed for the distribution of goods: a modification of Rawls's (1971) difference principle applied to individual allocations (rather than to the basic structure of society, as Rawls intended); utilitarianism; and economic efficiency. By the modified difference principle, goods are distributed to the least advantaged group of people. By utilitarianism, goods are distributed so as to maximize total utility. By economic efficiency, goods are distributed so as to maximize wealth, since, by maximizing wealth, the winners in any efficient reallocation could compensate the losers and still be better off than before. If you are a member of the least favored class, it is to your advantage to try to convince others to follow the difference principle. If you are rich, it is probably to your advantage to convince others to favor efficiency. This is because the efficiency principle will sometimes lead to increases in your wealth at the expense of those less fortunate. The difference principle will clearly rule out such changes. Utilitarianism will also rule out some of them because the decrease in wealth that the poor must suffer so that you can gain will often mean more to them than your increase in wealth means to you. This is because the utility of wealth is typically marginally declining.
If you are completely uncertain about who you are, it is arguably to your advantage to endorse utilitarianism. And I should heed you if I don't know who I am. But this is not a good argument for utilitarianism, for we are not completely uncertain about who we are. You cannot appeal to my self-interest (broadly construed) to accept a rule when the self-interest that you appeal to is counterfactual (my not knowing my position in society) and when my true self-interest tells me that a different rule would be better. This is just an example of a more general problem with the appeal to self-interest as a way of persuading people to accept moral rules. Such appeals can lead to even more blatant forms of parochialism, such as racism or nationalism.
Altruism can remove the conflict between alternative norms that arises from self-interest. If I endorse the same rule that I would choose if I did not know my identity, then I cannot be suspected of this kind of selfishness, and such suspicion will not be an impediment to agreement. Note that the rule I would choose if I did not know my identity is also the rule that I would choose if I were perfectly altruistic (including myself only as one person within the scope of altruism), or if I were unaffected by the decision to be made. Here my reasons for norm endorsement come from altruism, defined as my goal for the achievement of the goals of others. If others know that my reasons for endorsing norms are altruistic, then they have another reason to accept my endorsement: their own altruism would lead to the same endorsement (if we agree on the facts, at any rate).
One altruistic decision rule for several people is the utilitarian rule that counts all of their goals equally. I will not here try to defend this rule, as distinct from some rule that favors some at the expense of others in an arbitrary way (not based on self-interest). However, the appeal to altruism as the basis of norm endorsement does remove an impediment to acceptance of the utilitarian rule, namely, the fact (argued above) that self-interest can lead to endorsement of conflicting principles such as efficiency and the difference principle.

Which goals or values should we count

The norm-endorsement view presented here is not just a way of getting to the same old conclusions that other utilitarian writers have reached. It provides answers to several questions about which goals or values should be counted in a consequentialist analysis. Sometimes these answers are different from those of other utilitarian approaches. The question of what should count is answered by asking whether our altruism (and self-interest) give us reason to endorse counting some particular kind of goal. Our altruism depends on someone's self-interest, so, ultimately, the question is whether endorsement of counting a particular goal serves people's self-interest.
In the rest of the paper, I provide some brief examples of how the norm-endorsement view can answer questions about what goals or values should count in a utilitarian calculus. These are the main questions where the norm-endorsement view differs from other utilitarian views.

Erroneous subgoals

Some goals owe their existence to other goals. For example, Helen has a strong liking for avocados because she believes that they are healthy and non-caloric, although she does not much like their taste. Her desire for avocados is therefore dependent on her desire to be healthy and slim. Such erroneous subgoals are dependent on false belief. True subgoals, and, more generally, true goals, are not. (Keeney, 1992, distinguishes between fundamental values, which are independent of beliefs, and proxy values, or subgoals, which are related to fundamental values through beliefs.)
Suppose you know that avocados are full of calories and saturated fat. Should you give her avocados as a present? (You do not know her well enough to point out her error.) To answer this question, let us apply the norm-endorsement approach. We all have an interest in supporting a general principle of helping each other to achieve our goals. This certainly applies to fundamental goals, those that are unaffected by changed beliefs about the extent to which their achievement promotes the achievement of other goals. But does this interest extend to subgoals based on error?
At first, we might think so. Putting ourselves in Helen's position, we might think of the pleasure we would experience at getting a nicely wrapped box of avocados. And surely the pleasure would provide a reason to give Helen the avocados. But suppose that our alternative gift idea would provide just much pleasure while being neutral toward other goals (e.g., a bouquet of flowers). In both cases, of course, Helen has a goal of getting pleasure itself, and this goal is equally satisfied, let us assume. Now we can face the issue of whether the avocados are to be preferred because Helen thinks they help her achieve her other goals.
In this case, it clear that we should endorse a rule that favors taking account of fundamental goals, not erroneous subgoals. If a subgoal is based on a false belief, then we would want someone to honor our true goals. It is these true goals that give us the ultimate reasons for what we endorse for other to do. The erroneous goals do not follow, so the chain of reasons is broken. By this rule, you should give her the flowers. (Perhaps criticizing the epistemic bases of our subgoals is what Brandt, 1988, means by "cognitive psychotherapy.")
Note that this criterion can be applied to the absence of goals as well as their presence. If a goal would be present except for an erroneous belief, then we should assume that it is present. Erroneous beliefs can include lack of knowledge of the existence of things. Thus, by this view, we can help people by giving them things that they "did not know they wanted." This possibility helps to avoid a common criticism of other forms of utilitarianism, namely, that it is unfair to those who are ignorant or whose imagination is limited (Elster, 1983).
The point here is that we would want each other to act on the basis of the truth. In practice, this requirement leads to all sorts of complexities, for people are rarely sure of their beliefs. Helen may be just as sure of her belief about avocados as you are of yours. In this case, she may want you to act on her belief. More generally, the principle of ignoring erroneous subgoals can lead to excessive paternalism. We sometimes think we are ignoring such goals, when in fact we are the ones in error. But this is a practical, prescriptive, issue, not a normative one. We must put this issue aside when asking whether have any reason to neglect erroneous subgoals.
I do not mean to minimize the ambiguities created in practice by the question of when a subgoal is erroneous. Those who attempt to impose ideals on others - for example, religious ideals - often argue that the goals that they frustrate are not true goals at all. On the other hand, sometimes people really do not know what helps them achieve their true goals, and we do well to put aside their opinions in making decisions on their behalf. To complicate matters further, people may have a goal of pursuing their own goals (true or false), that is, a goal of being autonomous. This goal, or its strength, may itself be erroneous, but it, too, must by respected to the extent to which it is real. I doubt that any single, simple principle can help us to make the truly difficult decisions that arise from this conflict.

Goals and preferences

I defined utility in terms of the achievement of goals rather than the satisfaction of preferences, although the latter formulation is a common one. In theory, we can measure goal achievement by asking people directly to estimate the extent to which a well-described outcome achieves their goals - separately or together. Although "preference satisfaction" can mean many things (including "goal achievement"), in its strictest sense it refers to a behavioral choice, a "revealed preference." If I choose an apple over a banana (or say that I would), then, in this sense, I prefer it.
Preferences in this sense often conflict with goals or values. People often make choices against their values because they lack crucial information (as I just discussed). Also, a great deal of psychological research (reviewed by Baron, 1994b) shows that people make choices that are inconsistent if we assume that they are trying to achieve a fixed set of goals. When preferences and values conflict, which should we count?
The norm-endorsement approach implies that we should honor values, not preferences. Our values are by definition the standards by which we evaluate all outcomes. We use these standards to criticize, among other things, our own preferences.

Future goals

Goals may refer to specific times, although they need not. A perfectly reasonable goal is "to get to the station before the train leaves." Another is "to have an increasing standard of living over my life." (The "principle of temporal good," discussed by Broome, 1991, ch. 11, is therefore not assumed here.) Goals can also change over time in their strength or existence. The time at which a goal is satisfied need not be the same time at which it is held.
Two questions arise about future goals. One is whether we should count them fully when we calculate total goal achievement. The other is whether (or when) we should try to bring future goals into existence. This section concerns only the first question.
Concern with the achievement of future goals is a bi-product of the impartiality of moral principles. Future people are people, so we should care about their goals, even if these goals are not yet present. The same goes for the future goals of a person who now exists. Each of us would want others to honor our goals, even if those others made their decisions before the goals existed (before they affected judgments), assuming, of course, that the others could anticipate the existence of these goals. Beyond this, people have present goals that their future goals are achieved, so future goals are relevant for achieving present goals as well. We would thus endorse norms for achievement of future goals (other things being equal).
This principle, together with the principles concerning erroneous subgoals, also obviate - in some cases - the need for an additional principle specifying that we should help others to achieve only their rational goals (e.g., Parfit, 1984, in the "Critical Present Aim Theory"). The goals that we consider "rational" for this purpose are often those that are not erroneous subgoals and do not conflict too much with future goals.

Goals for the unknown

Is it true that "What a person doesn't know can't hurt her?" If it is, we are justified in ignoring the wishes of the dead (except for the precedents they set for honoring other wishes) and in deceiving people to make them think that their goals have been achieved. The view that utility is a property of experience leads either to this attitude or to rejection of utilitarianism.
To determine whether goals for events not experienced are relevant by the norm-endorsement criterion, we ask whether we would endorse a rule or principle that these goals should be honored. If they are truly our goals, it is clear that we would. People pay money to lawyers to insure that their wishes will be carried out after they are dead. They pay detectives to find out if their spouses are cheating even when a positive answer can only make them unhappy. So they seem to have goals for events they do not experience. People who have such goals have reason to endorse norms that encourage others to honor these goals. People can, then, be hurt by their spouse's disloyalty, even if they never find out. Sleeping with someone else's spouse impairs the achievement of their goals in ways that they would not want, and they would want a norm that protects them from this.

Past goals

A related issue is the relevance of past goals for decisions made now that can affect their achievement. If someone had a goal concerning outcomes at some future time (e.g., their later life, or after their death), should we take this goal into account when the time comes, even though the goal itself is no longer present? If we ignore past goals, we need to draw a distinction between goals that people will have in the future - which I have argued should be considered - and past goals. Is it arbitrary to attend to future goals and ignore past ones?
In thinking about this issue, we must put aside many of the reasons that often cause us to honor past goals. We do honor people's wishes after they are dead, for example, but the reasons for this are often not intrinsic to the wishes themselves: Failure to execute someone's last will would undermine the incentive function of will making, which motivates people to work for their heirs as well as themselves. These issues concern future goals, the goals of those who will be helped by this work, for example.
Parfit (1984, ch. 8) points out that past and future goals often differ in relevant ways. Consider a future goal and a past goal, neither of which is now present - for example, a goal of a person who died and the goal of a person not yet born. Our decisions can affect the satisfaction of the future goal at the time the goal is present but they cannot affect the satisfaction of the past goal at the time it is present. If we limited our concerns to the satisfaction of goals when they were present, we would be able to distinguish past and future goals. We cannot limit ourselves in this way, however, because some goals really do concern some future time after the goals themselves are absent, and, most importantly, we might well want others to honor our goals in the future after our goals are psychologically absent. Because we sometimes want others to honor these goals, we would, other things being equal, endorse their inclusion.
Parfit (1984, p. 151) distinguish between desires concerning the future that are "conditional on their own persistence" and those that are not. The former apply only so long as they are present, and the latter apply even after they cease to exist. Unconditional goals give us reason to bind the future behavior of ourselves or others. If I desire now to go swimming tomorrow, and if tomorrow comes and I no longer have the desire, I have no reason to honor my former desire, because it was, from its inception, conditional on its own persistence. I could have stated initially, "I want to go swimming tomorrow, unless I change my mind." Other desires, such as those concerning my child's welfare after I die, are not conditional. I can speak meaningfully of the achievement of this goal being affected by events that occur after my death, and it is a goal that influences my present decisions: e.g., writing a will.
The difference between conditional and unconditional goals is this: For conditional goals, we prefer (or judge to be superior) options that would allow us to subvert the goals in the future, other things being equal. We have a goal of future freedom to change our mind. For unconditional goals, we have no general reason to judge options that preserve freedom as superior to options that do not. Our efforts to achieve these goals are not reduced if we know that we cannot undo the effects of our choices. We might even take steps to bind ourselves, to restrict our future freedom. Goals that we have for after our death are unconditional because we have them despite the fact that we cannot take them back.
We have reason to endorse the inclusion of both types of goals, just because they are both goals that people have. If we insist on honoring a past conditional goal that is no longer present, then we are subverting the holder's goal of being able to take back the goal. We would want others to honor our unconditional goals, however, whether they involved binding ourselves ("Don't give me more than three drinks even if I ask for a fourth") or desires for after our death (wills, etc.). We would want such a system even if no precedents were set concerning promise-keeping in general. It seems, then, that we should ignore past goals unless they are unconditional (on their own persistence). Past unconditional goals are those that each of us would want a fiduciary to honor. However, if the goal is conditional on its persistence and if it no longer exists, then it cannot be achieved (any better than it has been achieved already). It does not count, then, because it outside of our power to affect.
Parfit (1984, p. 157, also p. 151) is skeptical about this solution. He gives (among others) the example of his having wanted to be a poet when he was young. He thinks that his past desire gives him no reason to write poems now, even though the original desire was not conditional on its own persistence. He would exclude such goals. But how does he know that his desire to be a poet was unconditional (on its own persistence)? He might try to produce evidence of his saying "I want to be a poet even if it is not what I later want." I would argue that this is not enough unless we really believe that he would have been willing to limit his future freedom. It is doubtful that he would have done so. (If he would have done so, then he might be pursuing an erroneous subgoal).
Most of our desires (including Parfit's examples) are, I think, conditional. The major exception consists of those cases in which people anticipate a change and arranges matters so that their earlier desire will be fulfilled: Ulysses having himself bound to the mast; or women who ask that they give birth without anesthesia even if they should change their mind and ask for it during a normal birth. All statements of desire, it seems, have an implicit escape clause, "unless I change my mind," and we do not discourage people from changing their mind for their own good unless we are sure that this clause was crossed out. In general, then, desires are unconditional only when we would take steps to see that they are brought to bear on future decisions.

The rationality of goals

What makes goals rational? To answer this question, we can think of goals as consequences of decisions. Decisions affect goals in a variety of ways:

1. Most decisions create subgoals. If I decide to get a drink of water, I may have a subgoal of finding a drinking fountain.

2. Some decisions can be expected to bring certain goals into existence, strengthen them, or weaken them. If you decide to go to law school, you can expect to acquire at least some of the goals that lawyers typically have. If you undergo a religious conversion, in either direction, you can expect your goals to change drastically. Often we make decisions of this sort because we want our goals to change. Intentional goal change is also a means of self-control, as when someone cultivates a dislike of cigarettes.

3. When we bring sentient life into existence, or terminate life, we create or destroy goals. When we create life, we know only probabilistically what goals will result, but such uncertainty is not unique to this kind of decision.

The problem of the rationality and morality of goal change is therefore related to a variety of other problems, from the choice of one's life commitments to the question of population growth and abortion. What principles should we endorse for creation and destruction of 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, i.e., the achievement of (other) goals. Some of our goals concern goals themselves. We want goals that others approve of, goals that will bring good feelings when we try to achieve them, or goals that we are likely to achieve.
Decision about goals also affect directly the achievement of other goals that we already have. Typically, the addition of new personal goals impairs the achievement of other goals, but sometimes we are lucky and our efforts to achieve one goal help us achieve others as well. For example, in some companies, married men are trusted more than unmarried men, so an otherwise celibate workaholic might do better even in his work by adopting the goal of having a family.
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). We can evaluate such effects in terms of our current goals for the achievement of future goals. For example, I have considered running for congress, expecting not to win but to "educate" my fellow citizens. If I ran, my desire to win would increase as the race went on. In deciding whether to run, I must consider both the creation of this new goal and the (low) probability of its achievement. Because I do not think I could prevent this goal from developing, and because I have a present goal of not adding goals that are unachievable, I have decided not to run.
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 the best alternative option. 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. By this criterion, a new goal is rational if it meets this test.
Note that this method of evaluation by itself does not allow us to reason from any particular first principles. We cannot compare two sets of goals without some core set of common goals. In these respects, 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 wholes.

Goals of future people

I have argued that we should endorse caring about the goals of such future people as are assumed to exist (assuming that we include them in the scope of morality, an issue I have put aside). A different, more difficult, question is whether we should bring new goals into existence. The birth of a person will lead to the formation of many goals, which will be achieved to some degree, and which will affect the achievement of the goals of others. Adding a person, then, is like adding a goal within a single person. We are uncertain what the new goals will be, but we can still form an expectation about their degree of success and their effect on goals already present. The decision to add new people should, by this view, depend entirely on the extent to which they help achieve our present goals, including our goals for the creation of new goals.
Suppose that policy A will lead to the birth of certain future people, call them group A, and policy B will lead to group B. Group A is larger than B. In evaluating the policies, we should compare the goal achievement of the two groups in the usual way, as best we can, assuming that the group exists in each case. If average goal-achievement per person is constant, group A will have more goal achievement, so policy A would be favored on this ground alone. But we must also consider our own goals, and the goals of others, concerning the creation of new goals. Such goals could specify some optimal number of future people, with group B closer to the optimum than A. These goals could tilt the decision toward policy B. Possibly, a person with altruistic goals would want a sufficient number of future people to exist so that the long-term effects of her altruism could be maximized after her death. But the same goals would dictate that the population be sufficiently small so that those who exist could have good lives.
Against this view, Hare (1975) argued that, since each of us (presumably) would not have wanted not to have been born, then we must act toward others to see that they are born too, at least until the marginal utility of births is negative. Although the argument I gave for consequentialism was similar to Hare's (1963) argument for the Golden Rule, the rule he applies here, it was not the same. In particular, in my argument, our motivation to support moral principles came from our (collective) current goals. The unborn - those who are at issue when we make decisions about creating new people - do not have such goals (although they will if they are born). Consequentialism derived in this way need not be extended to those without goals (in the same way that it must be extended to future people conditionally on their birth). Rather, the creation of goals by creating people is analogous to the taking on of extra goals within an individual.
The question of bringing people into existence is dependent on what our current goals concerning future people would be if they were rationally adopted (i.e., consistent with other goals and not erroneous subgoals). This possibility gives us, potentially, a way out of Parfit's (1984) "repugnant conclusion" that an enormous population living in misery could be the best of all possible worlds.
I have assumed in this section that decisions about goals should be made independently of whose goals they are. The view of goals as independent of the people who have them is sometimes taken to imply that we can freely kill people so long as we replace them, a conclusion that is used to argue against the premise (Broome, 1985). This conclusion ignores, however, people's goals to achieve their present goals, and their goals for their lives as wholes. For these kinds of reasons, Singer (1979) argues that killing "persons" is worst than killing creatures whose goals are confined to their experiences of pleasure and pain.


Asch, D., Baron, J., Hershey, J. C., Kunreuther, H., Meszaros, J., Ritov, I., & Spranca, M. (1994). Determinants of resistance to pertussis vaccination. Medical Decision Making, 14, 118-123.
Baron, J. (1993). Morality and rational choice. Dordrecht: Kluwer.
Baron, J. (1994a). Nonconsequentialist decisions (with commentary and reply). Behavioral and Brain Sciences, 17, 1-42.
Baron, J. (1994b). Thinking and deciding (2nd ed.). New York: Cambridge University Press.
Brandt, R. B. (1979). A theory of the good and the right. Oxford: Clarendon Press.
Broome, J. (1985) The economic value of life. Economica, 52, 281-294.
Broome, J. (1991). Weighing goods: Equality, uncertainty and time. Oxford: Basil Blackwell.
Elster, J. (1983). Sour grapes: Studies of the subversion of rationality. New York: Cambridge University Press.
Gibbard, A. (1990). Wise choices, apt feelings: A theory of normative judgment. Cambridge, MA: Harvard University Press.
Hammond, P. H. (1988). Consequentialist foundations for expected utility. Theory and decision, 25, 25-78.
Hare, R. M. (1963). Freedom and reason. Oxford: Oxford University Press (Clarendon Press).
Hare, R. M. (1975). Abortion and the golden rule. Philosophy and public affairs, 4, 201-222.
Hare, R. M. (1981). Moral thinking: Its levels, method and point. Oxford: Oxford University Press (Clarendon Press).
Keeney, R. L. (1992). Value-focused thinking: A path to creative decisionmaking. Cambridge, MA: Harvard University Press.
Kupperman, J. (1983). The foundations of morality. London: Allen & Unwin.
Parfit, D. (1984). Reasons and persons. Oxford: Oxford University Press (Clarendon Press).
Rawls, J. (1971). A theory of justice. Cambridge, MA: Harvard University Press.
Ritov, I., & Baron, J. (1990). Reluctance to vaccinate: omission bias and ambiguity. Journal of Behavioral Decision Making, 3, 263-277.
Sabini, J., & Silver, M. (1981). Moralities of everyday life. Oxford: Oxford University Press.
Simon, H. A. (1990). A mechanism for social selection and successful altruism. Science, 250, 1665-1668.
Singer, P. (1979). Practical ethics. Cambridge University Press.
Singer, P. (1982). The expanding circle: Ethics and sociobiology. New York: Farrar, Strauss & Giroux.


1This article is a substantial reworking of the arguments in Baron (1993), chs. 1-3. I think Samuel Freeman for comments. Send correspondence to Jonathan Baron, Department of Psychology, University of Pennsylvania, 3815 Walnut St., Philadelphia, PA 19104-6196, or (e-mail)

File translated from TEX by TTH, version 3.59.
On 3 Feb 2005, 07:16.