Jump to ContentJump to Main Navigation
The Community of AdvantageA Behavioural Economist's Defence of the Market$

Robert Sugden

Print publication date: 2018

Print ISBN-13: 9780198825142

Published to Oxford Scholarship Online: July 2018

DOI: 10.1093/oso/9780198825142.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see www.oxfordscholarship.com/page/privacy-policy).date: 19 July 2019



(p.83) 5 Opportunity
The Community of Advantage

Robert Sugden

Oxford University Press

Abstract and Keywords

Chapter 5 proposes using opportunity rather than preference-satisfaction as a normative criterion. I present an ‘Individual Opportunity Criterion’, which essentially states that expansions in any individual’s opportunity set should be treated as having value to that individual. I give a contractarian justification for this criterion under the assumption that individuals’ preferences are neoclassical, arguing against philosophers who claim that preference-satisfaction is not an acceptable normative criterion, even when preferences satisfy neoclassical assumptions. I then show that the contractarian justification of the Individual Opportunity Criterion extends to cases in which those assumptions do not hold, including cases in which an individual’s decisions are dynamically inconsistent.

Keywords:   opportunity, contractarianism, dynamic inconsistency, preference-satisfaction, Individual Opportunity Criterion

To follow the contractarian approach is to use a particular model of the role of the normative economist. In this model, the individual members of a society are considered as potential parties to voluntary agreements. The economist addresses these individuals together, recommending terms on which they might agree. The content of a recommendation is that a certain agreement is in the interests of each individual, as he or she perceives those interests. So a first essential is to have a criterion for determining whether or not a proposal is in a person’s interests. What this criterion should be is the topic of this chapter.

It is fundamental to the contractarian approach that the ultimate authority for judging what is in a person’s interests is that person himself. The economist’s job is to design proposals that individuals will want to accept. But part of that job is to explain the nature of the proposals and why it is in each individual’s interests to accept them. This requires that the economist has some representation of individuals’ interests that is independent of those individuals’ prior attitudes to any specific proposal. She needs to be able to say: ‘Having represented your interests in this way, I can show you that it is in your interest to agree to this proposal’. To fill in ‘this way’, we need a criterion of individual interest.

5.1 The Individual Opportunity Criterion

In thinking about what this criterion should be, we must keep in mind the uses to which it will be put. We are looking for a criterion that can be used by economists when they make recommendations about economic policies and institutions. To be useful for this purpose, it must satisfy certain design constraints.

One such constraint is that the criterion should be general with respect to individuals. That is, whatever formula it uses to represent an individual’s (p.84) interests should apply to individuals in general. It is simply impractical to have a separate criterion for every individual, tailored to his particular requirements. For similar reasons, the criterion should be general with respect to applications—that is, capable of being applied to a wide range of economic problems. Since the criterion is to be used in economics, often in conjunction with predictions derived with the help of theory, it needs to engage with economic theory. Thus, it should be framed in terms of concepts that, while being intelligible to the individuals on whose behalf it is to be used, can be processed in economic analysis. (To say this is not to be imperialist on behalf of economics. Normative economics is not the only kind of normative analysis there can be; it is just the kind that economists do.) And the criterion needs to be operational. By this, I mean that there must be some reasonably well-defined method for satisfying its informational requirements. Here we must remember that a contractarian economist has to be able to show individuals that her recommendations are in their interests, as they perceive them. It is not enough for her to assert this and to appeal to their trust in her superior judgement or expertise. So the method by which the informational requirements are satisfied must be transparent—that is, capable of being stated openly and laid open to public scrutiny. And, as far as possible, that method should be objective: it should not depend on subjective or contestable judgements.

Of course, the most fundamental requirement for a contractarian criterion is that it can be endorsed by the individuals to whom the economist’s recommendations are to be addressed. Each of those individuals must be able to recognize it as a representation of his interests, as he perceives them. Or, more precisely, he must be able to recognize it as a representation of what he wants to achieve from agreements with others, in the kinds of cases that economists make recommendations about: the word ‘interest’ is only a shorthand for that. Remember that a contractarian economist does not have to endorse the criterion as a representation of what she judges to be the individual’s best interests, or of what in her judgement the individual ought to want to achieve from agreements. She is his agent, not his guardian.

The idea that an individual endorses a particular criterion must be understood in relation to the design constraints that any criterion has to satisfy. In endorsing a criterion, the individual need not affirm that it gives a perfect representation of his interests in each and every possible case to which it might be applied. Rather, he accepts it as a satisfactory representation of his interests, given the design constraints.

The criterion I propose to use is that of opportunity. In its simplest terms, the idea is that it is in each individual’s interest to have more opportunity rather than less. An individual’s opportunities can be thought of as the set of options from which he can choose, options being described in such a way that they are mutually exclusive and jointly exhaustive. Such a set is an opportunity set. (p.85) (It is also known to economists as a ‘choice set’, ‘feasible set’, or ‘menu’.) For example, consider an air traveller who is offered a choice of onboard complimentary drinks. His opportunity set might be the set O = {water, orange juice, coffee, nothing}. A different airline might offer the set O′ = {water, orange juice, coffee, beer, nothing}. If ‘water’, ‘orange juice’, and ‘coffee’ mean the same in the two cases, O′ offers strictly more opportunity than O in the simple sense that O′ contains all the options that O does, and something in addition that the individual might conceivably want to choose. According to my criterion, it is in the traveller’s interest that the set of complementary drinks is O′ rather than O. I will make no attempt to compare opportunity sets that are not nested in this way.1

In the rest of this chapter I explain this Individual Opportunity Criterion in more detail, and defend the claim that each member of society can endorse it as representing his interests, as he perceives them.

5.2 Preference-satisfaction

As a first step in explaining and defending the Individual Opportunity Criterion, it is useful to consider how it relates to the criterion that has traditionally been used in normative economics—preference-satisfaction. Economists who have used the preference-satisfaction criterion have usually assumed that individuals have integrated preferences over all outcomes that are relevant to the problem at hand, and that individuals’ choices are consistent with those preferences. As I pointed out in Chapter 1, how far these assumptions correspond with reality is open to serious question. But for the purposes of this and Section 5.3, I will bracket out those doubts and suppose that the assumptions are satisfied. Given this supposition, one way of justifying the use of the Individual Opportunity Criterion is to say that the more opportunity a person has, the more effectively his preferences will be satisfied. I will present and defend this justification and then consider whether it can be extended to cases in which individuals do not act on integrated preferences.

Let me introduce an imaginary character: Norman the Neoclassical Agent. With respect to all decision problems that he could meet in his private life, Norman has preferences that satisfy all the postulates of the conventional theory of rational choice. That is, he has consistent preferences over all relevant consequences, and those preferences can be represented by a well-behaved utility function. He has consistent beliefs about all relevant states of the world, and those beliefs can be represented by subjective probabilities. In every decision problem, he chooses an action that maximizes the expected value of utility. Whatever opportunity set he faces, the act of choosing from it is costless to him. His preferences are stable over time and are independent of (p.86) irrelevant features of context or framing. His beliefs are stable too, except in so far as they are updated in the light of new information, as Bayesian probability theory requires. In line with an assumption that is implicit in most rational-choice theory, Norman’s preferences are consequentialist. That is, they are defined over objects that can be interpreted as the final outcomes of decisions, and not as properties of the procedures by which decisions are made.

Clearly, any expansion in Norman’s opportunity set maintains or increases the degree to which his preferences are satisfied. Take the case of the complimentary drinks. If Norman faces the opportunity set O = {water, orange juice, coffee, nothing}, he will choose an option in O that he weakly prefers to all the others. Suppose this is coffee. Now consider the effect of expanding his opportunity set to O′ = {water, orange juice, coffee, beer, nothing}. Now he will choose an option in O′ that he weakly prefers to all the others. If this is coffee too, there has been no change in preference-satisfaction. If it is another option that was also in O, say water, the only explanation consistent with Norman’s being a neoclassical agent is that he is indifferent between coffee and water. So again, there has been no change in preference-satisfaction. The remaining possibility is that Norman chooses beer from O′. Since he has chosen beer when coffee was available, either he is indifferent between beer and coffee, in which case preference-satisfaction is unchanged, or he strictly prefers beer to coffee, in which case preference-satisfaction has increased. Thus, if Norman endorses the preference-satisfaction criterion as a representation of his interests, he has no reason to object to any recommendation that satisfies the Individual Opportunity Criterion. (Sometimes that criterion will positively support proposals that Norman is merely indifferent about, but it never supports proposals than are contrary to his preferences.) So one way of justifying the Individual Opportunity Criterion to a neoclassical agent like Norman is to convince him of the merits of the preference-satisfaction criterion.

Assessed against my list of desirable properties, and given that we are dealing with a neoclassical agent, the preference-satisfaction criterion scores rather well. Since, by assumption, Norman has consistent preferences over all consequences that are relevant to the problems an economist might address, that criterion is extremely general. Preference is a readily intelligible concept, and it is central to economic theory. Since Norman’s preferences are revealed in his private choices, the informational requirements of the criterion can be met by observing those choices. By assumption, there are no inconsistencies between the preferences he reveals at different times, in different decision problems or in different contexts. So, if the criterion is accepted, identifying Norman’s interests—that is, identifying what he wants to achieve from agreements with others—is a technical problem in economics, in much the same way that discovering the height of a mountain is a technical problem in surveying.

(p.87) The most important question, of course, has still to be asked: Can the preference-satisfaction criterion be endorsed by Norman himself? In posing this question, let us take it as given that the domain of collective decision-making in which the contractarian economist makes recommendations is distinct from the domain of private life in which Norman’s choices have been assumed to reveal his preferences. Thus, the assumption that Norman acts on his preferences when making private decisions does not, as a matter of logical necessity, entail that he prefers it to be the case that the satisfaction of his preferences is used as the criterion of his interests for collective decision-making. Nevertheless, the latter proposition is very plausible.

It surely is a tautology that what a person chooses voluntarily is what, at the moment of choice and in the light of whatever deliberation he has engaged in, he most wants to choose. By assumption, Norman’s private choices reveal an integrated system of preferences. To use the satisfaction of those preferences as the criterion of Norman’s interests is to identify his interests with what, in all his private decisions, he most wants to choose. On the face of it, one might expect Norman to endorse this representation of his interests, at least for the purposes of arriving at and justifying economic recommendations.

Why might he not endorse it? If Norman is to reject the preference-satisfaction criterion, it seems that he must assert that what he consistently wants when he makes private choices is not what he wants to achieve from collective agreements. I can see two different ways of making sense of that assertion.

The first is to draw a distinction between private and collective decision-making. Norman might believe that some considerations that are relevant in one of these domains are not relevant in the other. For example, he might believe that the political community to which he belongs has a collective responsibility to promote particular forms of human excellence or to conserve distinctive features of its culture, and yet not attach any value to those matters in his private decision-making. This way of thinking about collective decision-making is not incoherent, but it is contrary to the spirit of the contractarian approach. As I explained in Chapter 3, this approach rests on a model of politics as negotiation. Politics is represented as a search for mutually beneficial agreements between individuals with given interests, not as public reasoning about what is socially valuable. I have already said what I can to persuade the reader of the virtues of the contractarian approach; I have nothing more to add here.

The second way in which one might understand how Norman might reject the preference-satisfaction criterion depends on a distinction, sometimes made by economists, between two personae or ‘selves’ of a given individual—the acting self and the reflective self. The acting self oversees acts of choice; the reflective self oversees the overall assessments that the person makes about his (p.88) life. By assumption, Norman has an entirely consistent acting self, whose choices reveal a well-articulated system of preferences. But it is still just about conceivable that the reflective Norman believes that those preferences, despite their internal consistency, do not reflect his interests. When he is thinking about his life as a whole, he might be convinced that his private choices are systematically distorted by errors of reasoning or failures of self-control, and as a result do not serve what (in his reflective moments) he regards as his true interests. And, despite recognizing this fact, he might believe that his future choices will be similarly flawed. (For example, the acting Norman might have consistent preferences with respect to smoking, revealed in his behaviour when he buys his daily supply of cigarettes, but the reflective Norman might consistently think that smoking was contrary to his true interests.) The reflective Norman might then have reservations about endorsing the preference-satisfaction criterion.

Even so, that is not to say that, all things considered, there is any better option. A criterion designed to implement individuals’ reflective judgements about their own interests would require some operational and transparent procedure for discovering those judgements, at least to some degree of approximation. A reflective Norman who thought carefully about the difficulties and risks of licensing any decision-making authority to act as the interpreter of his judgements about his interests might well decide that those interests were more reliably represented by the preferences of his acting self. I do not want to claim that there are no circumstances in which people, thinking reflectively about their best interests, are so conscious of (what they then see as) their error-proneness or weakness of will as choosers that they are willing to empower certain kinds of decision-makers to act as their guardians. But I submit that these cases are exceptional. I shall say more about them in Section 7.3, when I look at some of the implications of using opportunity as a normative criterion.

Any practicable form of normative economics depends on modelling simplifications. It would be unrealistic to expect to find a general, operational, and transparent criterion of individual interest that would be endorsed by every person in all circumstances. But were we able to assume that individuals act on integrated preferences, there would surely be a strong case for using preference-satisfaction as the criterion of individual interest. And that case would provide equally strong support for the Individual Opportunity Criterion.

5.3 ‘Mere’ Preferences?

Although most normative economists have used preference-satisfaction as the criterion of each individual’s interests, moral philosophers have often objected to this practice. They have argued that preference-satisfaction in itself (p.89) does not have normative significance. If the satisfaction of preferences has value (it is claimed), that must be because preferences are indicators of something else, something that does have value in itself. And then we can look for principles for screening out those preferences that are not indicators of value.

Typically, the principles that moral philosophers propose can be construed as conceptions of individual well-being, broadly conceived. The idea is that what really matters is each person’s well-being, and that preference-satisfaction matters just to the extent that it contributes to well-being. The philosophical literature provides a range of subtly-nuanced alternative conceptions of well-being, with corresponding screening principles. Some proposals appeal to informed desires, understood counterfactually as what a person would prefer for herself after ideally rational and informed deliberation (Arneson, 1989). Recall that Thaler and Sunstein (2008) use a similar criterion to determine when paternalistic interventions make choosers better off (see Section 4.1 above). Some proposals appeal to considered preferences, understood as preferences that are internally consistent and that are stable under experience and reflection (Gauthier, 1986). Some appeal to objective lists of the supposed components of human well-being, allowing that each individual may, within reasonable limits, weight these components in different ways (Griffin, 1986; Nussbaum, 2000). Some appeal to a reasoned consensus, reached through public discussion, about what human beings have reason to desire (Sen, 1999). And so on.

In denying that preference-satisfaction has normative value in itself, many philosophers argue that ‘mere’ preferences do not provide reasons for action. If a person is to reach a reasoned decision about what to do (it is said), she cannot treat her own preferences as data.

To understand this argument and how it relates to economics, one needs some knowledge of how choice is represented in the philosophy of mind. Philosophers make a distinction that is absent from conventional economics, between ‘desire’ and ‘intention’. To desire something is to have an inclination to choose it. That inclination is understood as a mental state that is in some sense external to the cognitive part of the mind that engages in reasoning. The cognitive mind is aware of the inclination and may treat its existence as a premise in its reasoning but, except in pathological cases, is not constrained by it. The cognitive mind’s reasoning leads to the formation of intentions. These are mental states that record decisions reached by the cognitive mind. Intentions can act as direct instructions to the acting parts of the body, or they may be stored to be used as premises by the cognitive mind in making more fine-grained decisions. For example, suppose I am in a pub, having just finished my first glass of beer. I am conscious of an urge to have to a second. That is a desire. Having thought about that urge and about other factors that bear on my decision problem, I decide to have a second beer. That is the formation of an intention. I then set about acting on that intention (walking to the bar, (p.90) engaging the bartender’s attention,…), making finer-grained decisions in doing so. After carrying out the intention, I experience the pleasurable sensations of drinking the second beer. Those sensations are conceptually distinct from the desires that prompted the decision that caused them.2

It is far from obvious how (or indeed if) the economic concept of preference fits into this framework. On one reading, favoured by some philosophical critics of the preference-satisfaction criterion, ‘preference’ is the comparative form of ‘desire’: to prefer x to y is to have a stronger desire for x than for y. This makes preference an input to the individual’s reasoning about what to choose. On another reading, ‘preference’ is the comparative form of ‘intention’: to prefer x to y is to have a settled intention to choose x rather than y. This makes preference an output of reasoning. On a third reading, favoured by Daniel Hausman (2012) and discussed in Section 4.5, a preference between two options is an all-things-considered comparative evaluation of them. For the moment, I will leave open the question of how the economic and philosophical frameworks map on to one another, and focus on what philosophers have actually said.

Consider the following passage from a philosophical critique of preference-satisfaction as a normative criterion. The author is Philip Pettit:

in deliberating our way to action we have to take our start, not from the fact that we desire certain goals, but from the fact that, as we see things, those goals are desirable or good or valuable or whatever. This line fits with our ordinary practice and with the long tradition of thinking that the major premise in a practical syllogism should not mention that some state of affairs is desired but rather the fact that is worthy of being desired.…Deliberation tries to track the true and the valuable, not the believed and the desired.

(Pettit, 2006: 144)

According to Pettit, a substantive theory of rational or reasonable choice must provide an account of the value-making properties that make some goals desirable and others not: hence the need for theories of well-being. Thus Pettit is able to conclude that ‘preference-satisfaction should not normally figure as a deliberative concern’, either when ‘individuals deliberate about what they ought individually to do’ or when ‘authorities or commentators deliberate about what good government ought to try to do for its people’ (2006: 131).

Pettit’s idea, as I understand it, can be illustrated by the following example. Suppose I am ordering a meal in a restaurant. There are two dishes on the menu, steak and salmon. Consider the following schema of practical reasoning (that is, reasoning about what to do), in which P1 and P2 are premises and C is the conclusion to be drawn from those premises:

  • Schema 1
  • (P1) I must choose steak or salmon.
  • (P2) I desire salmon more than I desire steak.
  • So (C) Let me choose salmon.

(p.91) Here ‘Let me…’ represents an intention to perform a particular action.

In this schema, a proposition about desire is being used as a premise in deliberating about what to do. Pettit would object to this as an unsatisfactory form of practical reasoning. It might be more accurate to say that, for Pettit, it is not reasoning at all. The schema describes a form of mental activity in which the cognitive mind moves straight from the recognition of a desire to the formation of an intention to act on that desire. According to Pettit, an individual whose behaviour is governed in this way is not an agent: ‘The fundamental tenet of our common sense psychology of human agents is that agency involves acting to realize various goals in a way that is sensible in light of the apparent facts’ (2006: 138). Pettit seems to be hinting that the abdication of agency, as exemplified by Schema 1, is unfitting for a human being.

Pettit presumably thinks that my reasoning in the restaurant ought to be more like this:

  • Schema 2
  • (P1) I must choose steak or salmon.
  • (P2) I would enjoy the salmon more than I would enjoy the steak.
  • (P3) Enjoyment is worthy of being desired.
  • So (C) Let me choose salmon.

In Schema 2, unlike Schema 1, none of the premises refers to my prior inclinations towards the actions I am deliberating about. P2 makes predictions about the mental states to which these actions will lead; P3 attaches value to those mental states. As Pettit would say, my deliberation about what to do is trying to track the true (how much I would in fact enjoy the two dishes) and the valuable (what truly is worthy of being desired).

Hausman (2012: 88) makes a similar argument, claiming that ‘satisfying preferences does not by itself contribute to welfare. Preference satisfaction theories of welfare are untenable’. Hausman is explicit that his argument does not depend on challenges to the conventional assumption that individuals have integrated preferences: it applies ‘even when preferences are informed, rational, and generally spruced up’ (2012: 77). Like Pettit, Hausman begins by looking at choice from the first-person perspective of the chooser. He claims that when a person thinks about what she should choose, she looks for reasons for doing one thing rather than another. Desires are not reasons in this sense:

From the first-person perspective of someone thinking about what to do, reasons are paramount. The question agents ask themselves is not ‘Given my beliefs and desires, what do I predict that I will do?’ but ‘What should I do’ or ‘What do I have most reason to do?’…To decide what to do, I try to be guided by what the facts are and by what is valuable.…That I happen to believe P and desire X is not decisive, because I can step back and question my beliefs and my desires. My awareness of a (p.92) desire to do X does not automatically incline me to do X intentionally unless I can see some reason to do X.

(Hausman, 2012: 5)

From the proposition that first-person reasoning about what to do should refer to reasons and not desires, it is a short step to an apparently similar proposition expressed in the third-person perspective of normative economics: that a criterion of what is in an individual’s interests should refer to what that individual has reason to choose, rather than to what he in fact desires or prefers. Thus, having asserted that ‘preference-satisfaction theories of well-being are mistaken’, Hausman immediately goes on to say:

Another way to support this conclusion is to ask what moral ‘pull’ satisfying someone’s preferences should have on others. In my view,…none at all. If something is valuable to people only because they want it, then their getting it has no direct moral importance for others. I have reason to help others to get what they want only if I can see how what they want is worth wanting, or why their lives will be in some way better if they get what they want.

(Hausman, 2012: 86)

Notice how, in a discussion about how to practise normative economics, Hausman has taken the standpoint of a disinterested philanthropist who gratuitously ‘helps others’. A similar standpoint is implicit in Pettit’s critique of the use of the preference-satisfaction criterion in deliberation ‘about what good government ought to try to do for its people’. If one imagines a wealthy and philanthropic private individual, leafing through unsolicited requests for money and deciding which of these she will accede to, requests that simply take the form ‘I want…’ do indeed seem out of place. The philanthropist is entitled to expect the writer of a begging letter to provide reasons why his wants are worthy of being met, rather than merely reporting his preferences. But is the relationship between a philanthropist and the objects of her charity the right model of the relationship between government and citizen?

Consider a different model. Suppose I have a friend who is confined to her house by illness. I offer to do her shopping for her, spending her money. She tells me what she wants to have bought, and I do my best to buy it. When I act on her behalf at the supermarket, her preferences surely are reasons for me. Suppose she has told me that she prefers whole milk to skimmed. Then, when I am at the supermarket, I choose whole milk because that is what she prefers. I don’t ask myself whether skimmed milk would be better for her, all things considered. That is a question that she might (or might not) want to ask herself when deliberating about her diet. But if I know what her preferences are, and if there is no ambiguity about what they imply with respect to what is on sale at the supermarket, it is not part of my role as her shopper to enquire into the reasoning that lies behind them.

I take this example to show that, when two people are in a principal–agent relationship, the preferences of the principal can be reasons for the actions of (p.93) the agent. And, more relevantly for the contractarian approach, the principal can want it to be the case that, when the agent is acting on her behalf, he treats her preferences as reasons. In other words, the principal can want the agent to use preference-satisfaction as the criterion of her interests. Pettit’s and Hausman’s arguments provide no objection to that proposition.

Here is a related example, which touches on questions about paternalism.3 Suppose I am the parent of a six-year-old child. It is early December and I am buying Christmas presents. I know that my child has a very intense desire for a certain toy. It is rather expensive, but can be bought within the budget that I have set aside for her presents. I see it as harmless but rather tacky, and predict that it will provide only brief excitement before she loses interest in it. There are alternative presents that I think would give her more pleasure over the coming year. Is her preference in itself a reason for me as parent, choosing what presents to give her? My own view (as author, and as one-time parent of young children) is that this is a significant reason, even though it is not necessarily decisive. The child’s preference has a moral pull on the parent that an unsolicited ‘I want…’ letter from a stranger does not. Why? The child is too young to control a budget sufficient to buy major items for herself, but she is still a full member of the family with preferences of her own, not a passive object of the parent’s benevolent concern. In buying toys for the child from the family budget, the parent’s relationship to the child has some of the characteristics of that of the shopper to the housebound friend. Even parents of young children should not be too paternalistic.

Now consider another way in which preferences might be thought of as reasons. Suppose that, instead of ordering a meal in a restaurant, I have decided that I want to buy a pre-prepared meal to eat at home. The next step is to choose which of two nearby shops, A and B, to go to in order to buy the meal. Both shops are operated by the same retail chain. The main difference between them is the number of products they stock. A is a convenience store, with a limited product range. B is a large supermarket which stocks everything that can be bought at A and much more besides. In keeping with the assumption I have maintained up to now in this chapter, suppose that my preferences over meals are stable and coherent. I know what my preferences are, but I do not know exactly which products the two shops stock. Still, I know that the opportunity set I will face if I go to B is a strict superset of the one I will face if I go to A. Is that a reason to go to B? It seems entirely reasonable to say that it is.

The most obvious way of reasoning to the conclusion that I should go to B is to use the criterion of preference-satisfaction. If my preferences are stable and coherent, and if I will act on those preferences whichever shop I go to, it is clear that the meal I end up buying will satisfy my preferences at least as well, and possibly better, if I go to B rather than A. (Compare the case of Norman (p.94) and the complimentary drinks, discussed in Section 5.2.) My reasoning can be represented as:

  • Schema 3
  • (P1) I must choose shop A or shop B.
  • (P2) Shop B will satisfy my preferences at least as well as, and possibly better than, shop A.
  • So (C) Let me choose shop B.

There is no circularity and no abdication of agency here. This schema treats the preferences I will act on in the shop I go to (whichever shop that turns out to be) as reasons for choosing between shops. But the reasoning is not circular: P2, the premise that refers to preferences, concerns preferences between meals, while the conclusion is an intention to choose between shops. There is nothing in Schema 3 that excludes the possibility that, when I get to the shop, my choices will be influenced by reasons, as Pettit and Hausman think they should be. But when I am choosing which shop to visit, I do not need to anticipate those reasons.

A critic might object that Schema 3 needs an additional premise, analogous with P3 in Schema 2, to the effect that preference-satisfaction is worthy of being desired. I disagree. I am using the concept of preference-satisfaction as it is normally used in economics. To say that, at the moment of choice in the shop, I prefer some product x to some other product y is to say that, at that moment, I want to choose x rather than y. If it then matters to me that what I choose is worthy of being desired, considerations of desire-worthiness will influence what I then want to choose, and so will be reflected in my preferences. If not, not. I am not abdicating agency; I am merely deferring the exercise of it. If there is an unstated premise in Schema 3, it has nothing to do with what is desire-worthy; it is something like ‘Let me be able to have what I want’.

In each of the three cases I have discussed—shopping for the housebound friend, choosing the child’s present, choosing between the two shops—preferences act as reasons for decisions; but the context in which preferences are revealed is distanced in some way from the problem for which they are treated as reasons. When I shop for my housebound friend, or when I choose a present for my child, another person’s preferences are reasons for a decision that I make. When I choose which shop to visit, my future preferences between products are reasons for my current choice between shops. In using preference-satisfaction as a decision criterion, I do not need to deny the possibility of asking whether preferences are based on good reasons. But that question is not relevant for the problem at hand. Thinking about reasons for the preferences that are to be satisfied can be left to another person (the friend, the child) or to another time (when I get to the shop).

(p.95) There is a similar distancing when preference-satisfaction is used as a criterion of individual interest in contractarian economics. The individual licenses the economist, when making her recommendations about collective actions, to use that criterion as a representation of what he (the individual) wants to achieve from those actions. Whether there are good reasons for those preferences is a matter for the individual himself; the economist can quite properly bracket out that question. Indeed, the individual might reasonably say that it is not the economist’s business to enquire into his reasons for wanting what he wants.

I must say, however, that I do not share Pettit’s and Hausman’s belief that, from the first-person perspective of the chooser, reasons must be paramount. I think Hausman is wrong when he says: ‘My awareness of a desire to do X does not automatically incline me to do X intentionally unless I can see some reason to do X.’ Nor do I agree with his suggestion that the question ‘What should I do?’ is equivalent to ‘What do I have most reason to do?’ As a matter of psychology, a desire for something surely is an inclination to choose it. I cannot see why one needs to look for reasons for doing what one desires to do, although of course one needs to keep a lookout for reasons for not doing so. And, in my experience, deliberation about what to do often calls for insight into one’s own feelings rather than the appraisal of reasons. (Suppose I am buying a new car and have narrowed down the choice to two models. One is more reliable and more economical, the other has cuter looks and is more fun to drive. I have desires pulling in both directions. It seems that I have to ask myself ‘What do I have more desire to do?’)

The relationship between reason and desire in choice is a deep topic in the philosophy of mind. Some philosophers, like Pettit and Hausman, give reason the primary role. Others, following David Hume (1739–40/1978), argue that when applied to decision problems, reason is inert unless it has desires to engage with, and that what a person desires is ultimately a matter of feeling. As an amateur philosopher, I side with Hume.4 For the purposes of this book, however, I do not need to take any particular position on these issues. The criterion of preference-satisfaction can be endorsed from either point of view.

Consider two more imaginary characters, Rachel the Reason-following Agent and Desirée the Desire-following Agent. For the moment, I assume that both of them have integrated preferences. Rachel thinks of her preferences as grounded on good reasons. Her preferences follow her desires only when she believes that what she desires is worthy of being desired. Desirée thinks of her preferences as grounded on her desires. Notice that nothing that I have said implies that Desirée’s preferences are less prudent or less moral than Rachel’s. To the extent that Rachel’s preferences are prudent or moral, she can attribute this to her recognition of reasons for not acting myopically or selfishly. To the extent that Desirée’s preferences have the same properties, (p.96) she can attribute this to whatever genetic, psychological, and social influences have made her desires what they are.

When Rachel endorses preference-satisfaction as a criterion of her interests, she is endorsing a criterion that identifies those interests with what, in her judgement, she has most reason to choose. When Desirée endorses preference-satisfaction as a criterion of her interests, she is endorsing a criterion that identifies her interests with what she in fact most desires to choose. Each can endorse the criterion, as applied to her, from her own philosophical standpoint. For a contractarian analysis, that is enough. It is not necessary that Rachel thinks that the satisfaction of Desirée’s preferences is a good criterion of Desirée’s interests, or vice versa. Rachel may think that Desirée is misguided in believing that the preference-satisfaction criterion represents her interests, when it ‘merely’ implements her desires. She may even think that Desirée’s attitude to choice shows that she is not an agent in the sense that she (Rachel) thinks fitting for a human being. But that is not how it looks to Desirée herself. Desirée (as I wish to represent her) does not think of her desires as alien forces that have somehow subverted the reasoning of her true self and led her to make unintended choices. She sees herself as an autonomous agent who does not need reasons in support of her desires. In so far as she can make any sense of the idea of a true self, she thinks that her desires are just as much part of that self as her ability to reason. If one takes the contractarian approach, there is no need to adjudicate between Rachel and Desirée. Neither has any standing to question the other’s judgement of her own interests.

5.4 Opportunity when Preferences are Liable to Change

So far, I have argued that a person with stable and well-articulated preferences can endorse the Individual Opportunity Criterion as a means of satisfying those preferences. I shall now argue that one of the advantages of using opportunity rather than preference-satisfaction as the criterion of individual interest is that it can deal more easily with changes of preference.

Here I am not referring to the kind of preference change that occurs as a rational response to new information, and which Bayesian decision theory can represent as an updating of beliefs. The cases I am concerned with are ones in which a person simply changes her mind about what she wants, given her beliefs about how the world is. Changes of mind may not fit easily into the framework of rational-choice theory, but they are surely one of the constants of human life.

Consider again the case in which I have to choose between two shops at which to buy a pre-prepared meal. I know that shop B stocks every product (p.97) that is stocked by shop A, and more besides. In the original version of the example, my preferences over products were stable, but I did not know exactly which products were stocked by each shop. The criterion of preference-satisfaction recommended me to go to B. But suppose instead that my preferences are not stable. Suppose I know that, between setting out for the shop and taking a chosen product to the check-out, I am liable to change my mind about what I want to eat. These changes of mind are not responses to new information. But that does not mean that (as viewed by me) they are the result of forces that compromise my autonomy as an agent. Perhaps, like Rachel, I always try to choose what I have most reason to choose, and my judgements about the relative weights of different reasons are liable to change. Or perhaps, like Desirée, I always act on my desires, and my desires are liable to change. Either way, I cannot choose to go to B as the best means of satisfying my given preferences: when I am choosing between shops, my preferences are not given. But it is still true that, whatever my preferences turn out to be, I will be better able to satisfy them at B than at A. And that still provides a strong reason in favour of going to B. Expressing this idea more generally, I do not need to have stable preferences in order to endorse the principle that my interests are served by my having a larger opportunity set. It is sufficient to say that I want to be able to satisfy my preferences, whatever they turn out to be, as fully as possible.

It might seem that this argument does not apply if, at the time that a person is choosing between alternative opportunity sets, he knows that his preferences over the options in those sets are liable to be affected by factors that will be determined by his choice between the sets. Continuing with the story of the two shops, suppose that both shops stock a particular kind of paella and a particular kind of pizza. A displays the paella in a way that, for me, is particularly effective in arousing a desire to eat it; B displays the pizza in a correspondingly attractive way. Were I to go to A, I would choose the paella; were I to go to B, I would choose the pizza. Given all this, is the fact that B offers more choice overall than A a reason for me to go to B?

It might be argued that if my preferences depend on how products are displayed in a shop, I cannot be acting as an autonomous chooser; I must be acting under the causal power of forces that are in some sense external to me. And if I know that this is the case, I cannot at the same time treat the non-chosen elements of my opportunity sets as genuine options for me. Thus, once I have decided which shop to go to, I have effectively decided to subject myself to forces that will result in my choosing paella (if I have gone to A) or pizza (if I have gone to B). So when I am choosing between the shops, it is as if that choice is between the opportunity sets {paella} and {pizza}, and the Individual Opportunity Criterion is silent on which of these better serves my interests.

(p.98) But remember that what is at issue is whether each individual can endorse the principle that it is in her own interests that she has more opportunity rather than less. So, in discussing autonomy, what matters is each individual’s understanding of her own preferences and choices, and not how these are viewed by anyone else. When an individual acts on context-dependent preferences, does she think of herself as the author of her own actions, or as the prisoner of a body that is responding to alien causal forces? Suppose I have gone to the shop that has the particularly attractive display of the pizza. As I take the pizza to the checkout, can I think of this as my choice, even though I realize that some of my desire to buy it has been triggered by the display?

I suggest that this question is ultimately about my sense of identity. If I define myself as an inner rational agent with an integrated set of neoclassical preferences, I might be troubled by the realization that my physical body seems to be behaving in ways that I cannot rationalize. But my guess is that most people do not think of themselves like this. If instead I think of my psychology as part of who I am, I can recognize that my choices are influenced by psychological mechanisms and still see them as mine. Why should I think of myself as lacking in autonomy because my choices are governed by mechanisms that operate through ordinary human psychology? Indeed, one might ask: What else could govern them? Once one thinks of identity in this way, it is no longer puzzling that a person might recognize that her preferences are context-dependent and still think it in her interests to have more opportunity rather than less.

The idea that opportunity has value for a person whose preferences are liable to change has implications for an influential critique of the preference-satisfaction criterion presented by Amartya Sen (1999: 58–63). This critique follows on from Sen’s discussion of alternative ‘informational bases’ for normative economics—the discussion that I commented on in Section 2.3. One of the informational bases considered by Sen is ‘utility’, interpreted as a representation of preferences. The core of Sen’s critique of the preference-satisfaction criterion is summed up in the following claim: ‘The mental metric of pleasure or desire is just too malleable to be a firm guide to deprivation and disadvantage’. Sen is particularly concerned with the mechanisms by which a person’s desires and capacities for pleasure adapt to unfavourable circumstances. Among his examples of people who can be affected in this way are ‘perennially oppressed minorities in intolerant communities, traditionally precarious sharecroppers living in a world of uncertainty, routinely overworked sweatshop employees in exploitative economic conditions, hopelessly subdued housewives in severely sexist cultures’. As a survival strategy, such persistently deprived people may ‘adjust their desires and expectations to what they unambitiously see as feasible’ (1999: 62–3).

(p.99) Sen wants to say that the utility-based approach can support unjust restrictions on the opportunities of people whose preferences have adapted to deprivation. Take the case of the hopelessly subdued woman in the sexist culture. The rules of her society debar her from activities that are seen as normal for men, such as taking paid work outside the home. Viewed in the perspective of the conventional preference-satisfaction criterion, whether that is a deprivation depends on her preferences. If she prefers her current activities in the home to paid employment, expanding her opportunity set to include paid employment would not increase her preference-satisfaction (but, of course, would not worsen it either). Sen’s response is that if that preference came about as her adaptation to the infeasibility of any other way of life, she is the victim of an injustice—whether or not she has any desire to change her situation. He proposes an approach to normative analysis in which ‘[t]he basic concern…is with our capability to lead the kind of lives we have reason to value’ (1999: 285). The subdued woman is the victim of injustice because she has reason to value a life that includes paid work, even if she does not recognize that value.

Because the Individual Opportunity Criterion does not take preferences as given, it is not vulnerable to Sen’s critique. According to that criterion, any expansion of a person’s opportunity set promotes her interests, irrespective of her actual preferences and independently of any argument about what, among the options that might be available to her, she has reason to prefer. The essential idea is that it is in each individual’s interest to have opportunities to satisfy not just those preferences that she currently has, but any preferences she might come to have.

This idea is central to John Stuart Mill’s The Subjection of Women (1869/1988), which I discussed briefly in Section 1.1. Mill is at least as conscious of adaptive desires as Sen is. He argues that the status of women in mid-Victorian Britain is a survival from the ‘primitive state of slavery’ that was once common for both sexes (1869/1988: 5–6, 31–5). In making this claim, Mill has to confront the fact that most mid-Victorian women do not perceive themselves as unfree. How can this be? Mill’s answer is that, because all but the most brutal men want their wives to be willing ‘favourites’ rather than forced slaves, women have been trained to develop those self-abnegating dispositions that are most attractive to, and most useful for, men. But, despite his belief that women’s desires are an adaptation to a form of slavery, Mill does not argue for the emancipation of women on the grounds that they have reason to desire a different form of life. Instead, he appeals to the principles of liberty and free trade. For example, responding to opponents who argue that it is essential for society that women marry and have children, and that this is proof that the natural vocation of a woman is that of wife and mother, Mill points out that goods and services that really are valued can always be supplied in the market, (p.100) so long as those who demand them are willing to pay the costs of supply. If men have a sufficiently strong desire for women to be housewives and mothers, the terms of freely-negotiated marriage partnerships will be such that women find it worthwhile to agree to them, given the opportunities available to them in a free labour market. In a free society, men will have to pay the price of the services they want, or go without (1869/1988: 28–30).

Notice that Mill does not need to address the question of how many women in fact desire a mode of life other than that of wives and mothers, or of whether women in general have reason to desire such a life. It is sufficient for his proposal that women and men should have the same opportunities with respect to marriage partnerships and labour contracts. Those opportunities are valuable to anyone who might ever want to use them, whatever their current preferences. The difference between Mill’s and Sen’s understandings of opportunity can be expressed in terms of the familiar problem of the ‘contented slave’ (which, according to Mill, is exactly the status of a happily married mid-Victorian woman). The problem is this: If slaves are contented, on what grounds can we condemn slavery? If our criterion is the satisfaction of actual desires, there seems to be nothing to criticize. Sen’s approach is to ask whether the contented slave has reason to desire a different way of life, from which she is debarred by her slavery. In contrast, Mill’s approach is to ask what would happen if the slave ceased to be contented: if she came to desire a different way of life, could she walk away from her existing one? Sen’s approach leads to a critique of slavery that can be opposed by those who claim that slaves have no reason to value a life outside slavery. In contrast, Mill’s argument retains its force against opponents who claim exactly this. If the slaves are so contented, why do they need to be prevented from walking away? And if the slaves really are contented, they can still recognize that their preferences might change.

5.5 The Continuing Person

So far, I have presented the Individual Opportunity Criterion as applying to decision problems that are faced at a single point in time.5 For example, in discussing the case of the two shops A and B, I used that criterion to compare the opportunity sets that the individual would face, having arrived at the respective shops. Let us call these sets OA and OB. However, in interpreting that comparison, I discussed the decision problem that the individual would face if choosing which of the two shops to go to. If, as I have assumed, the relevant features of the two shops can be described by the sets of options available at them, the choice between the shops can be represented as the opportunity set {OA, OB}, each of whose elements is itself an opportunity set. We might want to ask whether the (p.101) individual’s interests are better served by his having this choice than, say, by having no choice but to go to A, a possibility that can be represented as the singleton set {OA}. In this section, I consider how the Individual Opportunity Criterion might be generalized so that it applies to nested opportunity sets such as {OA, OB} and {OA}. This problem highlights some important issues about how the identity and agency of an individual should be understood if his preferences are not consistent over time. I will first present a generalization of the Individual Opportunity Criterion in abstract terms, and then explore its implications for a specific multi-stage decision problem.

As a starting point, consider two alternative options, v and w, which, if chosen by a given individual, would be experienced in some period t. (I will use the concept of ?periods? to refer to stages in a decision problem.) I leave open whether these options are opportunity sets in their own right. I introduce a formal concept of weak dominance among options; the proposition ‘option v weakly dominates option w in period t’ is written as vt w. The intuitive idea is that if v weakly dominates w, the opportunity provided by v is unambiguously at least as great as that provided by w. Here, ‘unambiguous’ is to be understood without reference to the actual preferences of the individual we are considering: to say that v weakly dominates w is to say that whatever the individual’s preferences (perhaps within some minimal bounds of reasonableness or plausibility), those preferences can be satisfied at least as well by v as by w. If v weakly dominates w but w does not weakly dominate v, I will say that v strictly dominates w, written as vt w. If each of v and w weakly dominates the other, I will say that they are dominance-equivalent, written as v ~t w; the interpretation is that the individual’s preferences, whatever they may be, can be satisfied equally well by v or w. I stipulate that weak dominance is reflexive—that is, vt v (and hence v ~t v) for all options v that can be faced in period t. However, I do not require that this relation is complete. If neither of v and w weakly dominates the other, I will say that they are not dominance-comparable, written as v #t w.

Given a weak dominance relation among options that might be experienced in period t, we might look for an analogous relation among the opportunity sets of which they can be elements. I propose the following dominance extension principle. Let V and W be alternative opportunity sets of this kind, and let t−1 denote the period in which those opportunity sets would be faced (and in which an option would be chosen from them). V weakly dominates W if, for every option w′ in W, there is some option v′ in V such v′ weakly dominates w′ in period t. I will write this as Vt–1W. The intuitive idea is that there can be no loss of opportunity in facing V rather than W if, for each option w ′ that can be chosen in W there is an option v′ in V which unambiguously provides at least as much opportunity. The relations of strict dominance, dominance equivalence, and dominance (p.102) non-comparability are defined from weak dominance in the same way as before. These definitions imply that reflexivity is preserved—that is, that for all opportunity sets V that can be faced in period t−1, V~t–1V. Thus, given a weak dominance relation among options than can be experienced in any period t, the dominance extension principle induces a dominance relation among the opportunity sets themselves. This principle can be applied repeatedly to induce dominance relations at successively higher levels of nesting (or, equivalently, in successively earlier periods).

This set of definitions can be interpreted as a generalization of the Individual Opportunity Criterion, as that was presented in Section 5.1. Take the case of the air traveller for whom opportunity sets are sets of alternative complimentary drinks. The period in which a drink might be consumed is t; the period in which an opportunity set is faced is t−1. Compare the sets O = {water, orange juice, coffee, nothing} and O′ = {water, orange juice, coffee, beer, nothing}. If we take it that any one of the five options (that is, the four drinks and ‘nothing’) might plausibly be preferred to all of the others, every pair of distinct options is dominance non-comparable. Thus, for every option x in O there is an option (the same x) in O′ which weakly dominates it, but the converse is not true; and so O′ strictly dominates O (i.e. O′≻t−1 O). If we read ‘strictly dominates’ as equivalent to ‘provides strictly more opportunity than’, we have reproduced the implication of the Individual Opportunity Criterion.

A model of a specific multi-stage decision problem may help to explain the intuitions behind this rather abstract analysis. The story behind the model is that the agent, Jane, is offered a rare opportunity to buy a ticket to attend some important sporting event. Going to the event would satisfy a long-felt desire and would give her intense but brief enjoyment. However, the ticket is expensive; buying it would require painful economies in other areas of her life. I begin with the version of the model in which Jane has the greatest freedom of action. There are four periods. Jane enters period 1 with an endowment m of money. In this period she has two options, buy and don’t buy. If she chooses don’t buy, she keeps her endowment and has no decision to make in period 2. If instead she chooses buy, she pays a price p1 (less than m) and receives the ticket. Then in period 2 she has two more options, hold and return. If she chooses hold, she keeps the ticket and what is left of her money. If she chooses return, she returns the ticket and receives an amount of money p2 that is slightly less than she originally paid. In period 3, there are no further opportunities to buy or sell tickets. If she has chosen don’t buy, her opportunity set in period 3 contains the various ways she could spend her original endowment m. Let this set be X. If she has chosen buy and hold, her opportunity set in period 3 is such that she can choose whether or not to go to the event and also how to spend mp1, the relatively small amount of money she still holds. Let this set be Y. If she has chosen buy and return, her opportunity (p.103) set in period 3 contains the various ways in which she could spend mp1 + p2. Let this set be Z. Since the only difference between X and Z is the amount of money that Jane is able to spend on things other than going to the event, I assume that Z is a strict subset of X.6 In period 4, she experiences whichever option she chose in period 3. All of this is known to Jane at the start of period 1.

Jane’s entire decision problem is described by the nested set S = {{X}, {Y, Z}}. The outer pair of curly brackets defines the choice facing Jane in period 1. This is a choice between the elements of the set that is specified by that pair of brackets—that is, the elements {X}, corresponding with don’t buy, and {Y, Z}, corresponding with buy. Each of these elements is a (possibly degenerate) choice problem that will be confronted in period 2 if the relevant action is chosen in period 1. The singleton {X} represents the fact that, if don’t buy is chosen in period 1, there is no choice to be made in period 2, and the options that will be available in period 3 are the elements of X. The set {Y, Z} represents the fact that, if buy is chosen in period 1, there will be a further choice to be made in period 2, between one action (hold) which leads to the period 3 opportunity set Y, and another action (return) which leads to Z. In this notation, each matched pair of curly brackets is associated with a specific period; the succession of periods is represented by successively deeper nesting of sets.7

Notice that if Jane acts on a single set of integrated preferences throughout the four periods (and knows that this is how she acts), she has no reason to choose buy followed by return. Whatever those preferences are, she can satisfy them at least as well (and typically better) by choosing don’t buy. This raises the question of whether the opportunity to choose buy followed by return is of any value to her. This question can be posed more precisely by considering a variant of the model in which the option of returning the ticket is not available: there are choices to be made only in periods 1 and 3. In this variant model, Jane’s entire decision problem is described by the nested set S′ = {{X}, {Y}}. How should the opportunities offered by S and S′ be ranked?

As a first step in answering this question, consider the three alternative opportunity sets, X, Y, and Z, that Jane might face in period 3. If these sets are considered independently of any preceding choices, they are the kind of objects to which the Individual Opportunity Criterion, as defined in Section 5.1, can apply. Since Z is a strict subset of X, we have X3 Z. Since neither of X and Y is a weak subset of the other, we have X #3 Y. Similarly, since neither of Y and Z is a weak subset of the other, we have Y #3 Z. Now consider the opportunity sets {X}, {Y}, and {Y, Z} that Jane might face in period 2. Applying the dominance extension principle, we have {X} #2 {Y}, {X} #2 {Y, Z}, and {Y, Z} ≻2 {Y}. Finally, consider the alternative decision problems that Jane might face in period 1. Applying the dominance extension principle again, we have {{X}, {Y, Z}} ≻1 {{X}, {Y}}. So the answer to the original question is that S gives more opportunity than S′: being able to choose buy (p.104) followed by return counts as an addition to Jane’s opportunities, even though the combined effect of the two actions is an unambiguous loss.

Why does this count as an additional opportunity? Think about how, facing the decision problem S, Jane might choose buy in period 1 and return in period 2. Suppose that, as I suggest is entirely plausible, there is no uniquely correct answer to the question of how much Jane should be willing to pay for the ticket. She can predict both the pleasures she will enjoy if she goes to the event and the sacrifices she will have to make in order to pay for the ticket, but she knows of no objective way of making these effects commensurable. In some states of mind, she finds herself giving particular attention to the pleasures of the event, and the price p1 seems worth paying; in other states of mind, when she attends more to the sacrifices, the ticket does not seem to be worth as much as p2. In period 1, she has to choose between {X} and {Y, Z}. Both options are eligible choices for her. Suppose she is then in the first state of mind, and doesn’t expect this to change. She buys the ticket. In period 2, she has to choose between {Y} and {Z}. Both of these options are eligible too: the fact that Z is inferior to an option she previously rejected does not make Z a bad choice now. Suppose that in period 2, she is the second state of mind, and returns the ticket. In neither period has she made an irrational decision; she has simply changed her mind. The added value of S relative to S′ is that it gives Jane the opportunity to act on this change of mind. My claim is that it is Jane’s interest to have this additional opportunity.

Some readers may want to object that, since Jane does not have a consistent set of preferences, it is a mistake to treat her as a single agent with definable interests. When representing individuals whose preferences change over time, decision theorists often use multiple-self models in which a person’s agency is broken down into that of two or more selves, each with its own integrated preferences; at different times or in different situations, different selves are in control of the individual’s actions. These selves are then treated as if they were distinct agents. In some models, the different selves of a single person interact strategically with one another, in the same way that players interact with one another in game theory. Sometimes, normative analysis privileges the preferences of one self, treating these as the ‘true’ preferences of the whole person. Sometimes, the preferences of the different selves are given equal moral status, and the whole person is treated as if she were a society with the individual selves as members. Sometimes, the person is assumed to have a coherent higher-level metapreference ranking of the various lower-level preference orderings which govern her day-to-day behaviour; normative authority is then located in each person’s supposed higher moral self.8

Applying this approach to the case of Jane, we might distinguish between an ‘experience-seeking’ self (Jane1, the self whose preferences are acted on in period 1) which is willing to pay a high price to go to the event, and a ‘prudent’ self (Jane2, the self whose preferences are acted on in period 2) which is willing (p.105) to pay only a low price. We might conclude that Jane1’s preferences would be better satisfied if the decision problem were S′, while Jane2’s preferences would be better satisfied if it were S. And then we (as impartial spectators or social planners) might try to decide which of these selves is truly Jane.

This modelling strategy, like that of the inner rational agent (to which it is closely related), often looks suspiciously like an attempt to continue to use rational-choice theory while describing the behaviour of people whose actual choices are inconsistent with that theory. One of its least attractive features is that it cannot recognize the continuing identity and agency of ordinary human beings who happen to choose in ways that disconfirm received theory. A failure of the theory is being re-cast as a failure of the individuals whose behaviour the theory is supposed to explain. As I explained in Chapters 3 and 4, this strategy opens the way to kinds of paternalism that are out of bounds to a contractarian.

I think we need a radically different conception of the continuing person. We should think of the continuing Jane—let us call her Jane*—as the composition of the selves which perform the various parts of whatever sequence of actions is in fact performed. For theorists who insist on modelling identity in terms of some kind of preference relation, this idea may seem strange. But, viewed from outside the framework of decision theory, it is a very natural way of thinking of identity. The continuing Jane* is just whatever Jane the human being is over time. What the continuing Jane* does is just whatever Jane does over time; what the continuing Jane* values is just whatever Jane values over time. In this perspective, it becomes clear how the continuing person can value the absence of constraints on her present and future actions.

Think again about Jane, facing the decision problem S. In period 1, Jane1 wants to choose, and does choose, buy. Since Jane* is Jane1 at this moment, it is also true that Jane* wants to choose, and does choose, buy. In period 2, Jane2 wants to choose, and does choose, return. Since Jane* is Jane2 at this moment, it is also true that Jane* wants to choose, and does choose, return. So Jane* wants and chooses to buy in period 1, and wants and chooses to return in period 2. In allowing this sequence of actions, S gives Jane* an opportunity to do something that she wants to do. If Jane* values opportunities to do as she wants, this feature of S has value for her.

Consider how, at the end of period 3, Jane (the human being) might reflect on the actions she has taken. A conventional decision theorist might point out to her that she has acted on preferences that are inconsistent over time, and that in consequence she has incurred an unambiguous loss—she has bought dear and sold cheap. Jane can concede this, yet still see both buying and selling as her autonomously chosen actions: she wanted to buy, and she bought; she wanted to sell, and she sold. She does not have to disown either of those actions as the work of an alien self, or as the result of weakness of will. (p.106) She can say that she has done what she wanted to do, when she wanted to do it. And she can say that it was in her interests that she was able to do this.

5.6 Responsibility

My proposal is that the continuing agency of a person across time and across contexts should be understood as the continuing existence of a self-acknowledged locus of responsibility. The intuitive idea is that a person is a continuing locus of responsibility—for short, a responsible agent—to the extent that, at each moment, she identifies with her own actions, past, present, and future. A responsible agent treats her past actions as her own, whether or not they were what she now desires them to have been. She treats her future actions as her own, even if she does not yet know what they will be, and whether or not she expects them to be what she now desires them to be.9

This conception of responsibility provides philosophical underpinning for the claim that opportunity has value. Consider the set of opportunities that are open to some individual across time. Is it a good thing that this set is larger rather than smaller? In conventional welfare economics, more opportunity is better than less only to the extent that it allows the individual to achieve a more preferred outcome; if the individual lacks integrated preferences, there seems to be no way of answering the question. In a model of multiple selves, the question has to be posed separately in relation to each self, and increases in lifetime opportunity may be judged to be good from the viewpoint of one self and bad from that of another. If appeal is made to metarankings, increases in the opportunities of selves which act on ‘inferior’ preferences may be judged to have negative value. But if an individual is understood as a continuing locus of responsibility, any increase in that individual’s opportunity is good for her in an unambiguous sense. The more opportunities she has, the more she—construed as responsible agent with a continuing existence through time—is free to do. This is true whether or not her actions across time are consistent with any one set of coherent preferences.

Benjamin Disraeli is supposed to have said, apropos of the attacks that any leading British politician faces in Parliament, ‘Never complain and never explain’.10 This aristocratic sentiment captures something of what it is to be a responsible agent. In relation to matters that affect only himself, the responsible agent asks of government only that it ensures him as wide a range of opportunities as possible. How he uses those opportunities is up to him, and he accepts sole responsibility for the consequences. He has no need to explain the decisions he has made, because they were no one else’s business. And because they were his decisions, he can have no complaint against anyone else about how they turn out.


(1.) There is a large literature in economics and philosophy dealing with the question of how to rank opportunity sets in terms of the ‘quantity’ or ‘value’ of opportunity they provide. There is general agreement that if there are two sets O′and O such that O′is a strict superset of O, O′ should be ranked as at least as highly as O; but there is no consensus about how to rank other pairs of sets. At various times I have made proposals about how this might be done (e.g. Jones and Sugden, 1982; Sugden, 1998a), but these are not used in the contractarian criteria I present in this book.

(2.) There is a related distinction in neuroscience between ‘wanting’ (desire) and ‘liking’ (pleasure). The effects of addictive drugs are sometimes modelled as directly attaching ‘incentive salience’ (the perception of being wanted) to mental representations of events associated with drug-taking, short-circuiting the normal process of reinforcement learning by which an experience of pleasure induces a desire to repeat the action that caused it (Robinson and Berridge, 1993).

(3.) I have an ulterior motive in raising the issue of the account that parents should take of their children’s preferences. In Section 3.4 I explained ‘why a contractarian cannot be a paternalist’, but added a footnote recognizing that some of the interests of children have to be looked after by guardians or trustees. Supporters of paternalism sometimes argue that, if one recognizes that parents can properly act as guardians of their children, it is dogmatic to deny that there can be analogous circumstances in which governments can properly act as guardians of citizens. It is at least a partial answer to this objection to say that even parents should not always be paternalistic: contractarianism is sometimes appropriate for parents. Compare John Stuart Mill’s ideas about the moral rules that mankind should practise in the family, discussed in Section 1.1.

(4.) For my reading of Hume’s decision theory, and my reasons for agreeing with its main features, see Sugden (2006a, 2008).

(5.) This section includes material from Sugden (2006b).

(6.) In effect, I am assuming that Jane can freely dispose of any money that is surplus to her requirements. Thus, any spending plan that is compatible with Z is also compatible with X.

(7.) Because the symbols X, Y, and Z represent sets of options that might be faced in period 3, each of these symbols is nested within two pairs of brackets (i.e. two pairs fewer than the number of periods). An equivalent representation of Jane’s problem in period 1 would write out the contents of these sets explicitly, replacing X, Y, and Z by {x1, x2,…},{y1, y2,…}, and {z1, z2,…} respectively. Then each option that might be experienced in period 4 would be nested within three pairs of brackets.

(8.) The idea of multiple selves as a modelling technique can be traced back to Strotz (1955–56), that of metarankings to Frankfurt (1971) and Sen (1977).

(9.) Schubert (2015) is one of several readers of my previous work who have criticized me for seeming to suggest that imposing constraints on one’s own future opportunities is irresponsible in the everyday sense of that word. I do not want to say that. Although I believe that economists and philosophers often over-state the (p.287) extent to which people want to constrain themselves, I accept that there are exceptional cases in which choosing self-constraint is a sign of responsibility. A classic case is the driver in the pub who, knowing that alcohol affects both his ability to drive safely and his ability to judge that ability, gives his car keys to a friend with the instruction not to return them if he drinks more than some limit. I will say more about such cases in Section 7.3.

(10.) The applicability of this quotation to my account of responsibility was first noticed by Schubert (2015). The quotation is recorded in the Oxford Dictionary of Political Quotations (Second edition, 2001: 115).