The bent of utilitarianism has been to do away with qualitative distinctions of worth on the grounds that they represent confused perceptions of the real bases of our preferences which are quantitative. The hope has been that once we have done away with strong evaluation, we will be able to calculate.

– Charles Taylor, “What is human agency?”

In the previous chapter, I argued that the structure of the problem of choice is incompatible with the Principle of Optimality underlying the utility framework. Decision-making, after all, is necessarily a twofold process: (i) the construction of a model that captures the uncertain preferences of the decision-maker and external events; and (ii) the use of this model in actually making choices. And with many unique or critical decisions it is this first stage that is not only interesting but also vitally important. With the utility framework, however, this is not the case, as the framework focuses solely on the decision process that takes place once future preferences and future events have been determined. And it is this determinateness that gives rise to difficulties in the utility framework, as the structural feature of decisions being made in a stream of time with an uncertain future is ignored.

In this chapter, the difficulties raised focus on the relationship between individual choice and the requirement that all outcomes must be reducible to a single utility metric. This requirement is expressed in the utility framework by both the principles of Sufficiency of Preferences and Commensurability. The Principle of Sufficiency of Preferences ensures that an individual’s given preferences are all that need to be considered when explaining choice, while the Principle of Commensurability ensures that the trade-off between competing preferences is always possible and the utility criterion can render a decision. In the following pages, I consider the empirically self-evident fact that decision-makers are self-conscious and possess volition and I explore the implications this has for the utility framework. I hope to demonstrate that the rigid utility relationship among outcomes, which is required by the utility framework to ensure commensurability, breaks down for decision-makers endowed with self-consciousness.

3.1 Incommensurability and Agency Both decision making and acting in the world require evaluation; we judge one state of affairs to be somehow better than another in order to decide or act. The issue of evaluation animates not only the utility framework, but also any approach to choice. (Recall Nussbaum’s [54] bagel example, examined in Section 1.4.)

If the world consisted of nothing save bagels and people, and people cared only about having as many bagels as possible, there would be no difficulty in distinguishing a more preferred outcome from a less preferred one. As long as we can count, our choice is easy. Because our utility function is exclusively concerned with a single variable, bagels, commensurability is assured – assuming, of course, that all bagels are alike. The world, however, does not co-operate in this regard. Instead, decision-makers are faced with multiple objectives without the means to fully attain them all, and trade-offs become necessary.1 In such a world, a mechanism is needed to evaluate the choice between two apples and three oranges or two oranges and three apples – and these comparisons between apples and oranges become an unfortunate fact of life.

As discussed at length in Section 1.4, it is as a solution to this dilemma of comparability that microeconomics2 (as a positive method for investigating decision making), ethical utilitarianism (as a way to investigate issues of individual morality and social policy), and decision theory (as an aid in complex decision situations) find their intuitive appeal. The advantages of the utility framework lie in being able to provide a determinate rule for evaluating the choice among competing alternatives. It is this aspect of determinateness and computability that appealed so strongly to Plato, Bentham, Sidgwick, and the economics profession – although Mill (the reader will recall) retreated slightly from this position.3

The key point here is that the decision-maker4 is viewed as merely the repository of a set of preferences faced with the computational task of choosing the best bundle of goods given the external constraints, a task which he can do more or less well. When choosing the maximal element, the decision-maker can only respond to the preferences which he finds resident within him. All the inputs to this decision process – both internal preferences and external events – are beyond the decision-maker’s control.

One might say that the preferences are completely constitutive of the individual’s identity. There is nothing else to the individual beyond these preferences (and perhaps a calculation unit that performs the mechanical exercise of actually determining the most preferred action given the fixed inputs).

Under these conditions, all values are immediately commensurable the individual does not have sufficient depth to do anything other than simply plumb his actual preferences. The commensurability so essential to the utility framework has been imposed on the individual by stripping him of all qualities other than his preferences.5 There is no room for will or volition.6 This chapter carves out an appropriate role for will, and demonstrates that this role is incompatible with the Principle of Commensurability underlying the utility framework. The role I develop views will as the faculty that must adjudicate among competing ways to impose commensurability on potential preferences. Note that this does not refer to competing ends but rather to competing sets of ends – that is, competing methods to combine preferences into a single preference scale that gives rise to decisions and actions.

The will, then, is not a call to action, but the prior mechanism that determines the relative weights of various preferences. If this role for the will exists, then the utility function is no longer completely constitutive of the individual – there must be other elements of the individual, beyond his preferences, that assist in determining which competing set of preferences will be acted upon. The existence of these other elements, we will see, raises complex issues well beyond those captured by the utility framework, including that of how the individual imposes commensurability on his preferences.

3.2 Actors and Agents  Although it is only a beginning, this section attempts to state more precisely the exact implications of taking agency seriously. A concept used in the philosophic literature, agency is the general capacity of self-conscious beings not only to choose one action over another (which is the condition fulfilled by mere, actors), but also to exhibit awareness of taking actions. Humans have this capacity of awareness or reflection, and this truth has some bearing on how we can appropriately model decisions we observe in the world.

To act in the world requires choosing among competing possible decisions. Choices are the consequence of judgments that certain outcomes are to be valued more highly than other possible outcomes. Judgment, then, is necessarily a process of differentiation as potential outcomes must somehow be valued differently to provide a basis for decision making. This does not imply that agents can never be indifferent to the various outcomes, only that decision making would not be possible if all outcomes were valued equally.7 All actors in the world, including many species other than man, have this capacity to draw these distinctions which furnish reasons for action. Without this ability, actors would have no mechanism for specifying the exact action that is to be chosen.8 Agents, however, are unique in having not only the capacity to choose desirable actions (with outcomes that are valued more highly than other potential outcomes), but also the awareness that such distinctions of value are being drawn. That is, agents can stand back and reflect upon these distinctions. “Self-awareness,” “self-reflection,” find “self- consciousness” are some of the terms used in the philosophical literature and everyday conversations to express this ability.

The ramifications of the distinction between actors and agents have been insufficiently understood by those drawing upon the utility framework. Were we merely actors lacking this self-awareness, then the distinguishing feature of human action (in contrast to that of animals) would be our capacity to calculate in order to secure what we desire. Elster [17:10] calls this ability “general, global maximizing.” Our goals would be more or less fixed (since we do not yet have sufficient depth to evaluate these goals, as the Principle of Sufficiency of Preferences implies), but the means to attainment would be under constant scrutiny. This common view of human decision making is an essential component of the utility framework, and Taylor [86:104] characterizes it well:

On this view, what is essential to the peculiarly human powers of evaluating and choosing is the clarity and complexity of the computation. Evaluation is assessment in light of our goals, which are seen ultimately as given, or perhaps as given for one part, and arbitrarily chosen. But in either case the evaluation process takes the goal as fixed. “Reason is and ought to be, the slave of the passions.”

Harsanyi [27:42] also sees our ability to calculate in the pursuit of our goals as the hallmark of rationality:

The concept of rational behaviour arises from the empirical fact that human behaviour is to a large extent goal-directed behaviour. Basically, rational behaviour is simply behaviour consistently pursuing some well defined goals, and pursuing them according to some well defined set of preferences or priorities.

The ability to perform highly complex manipulations – reason in the means-ends sense of the term – can, however, only be part of the description of an agent possessed of self-awareness. Indeed, Harsanyi’s “empirical fact” of goal-directed behavior seems to apply more to heat-seeking missiles than human beings. The distinguishing feature of agents’ behavior cannot only be the ability to pursue goals in some orderly fashion.

Rather, agents have the capacity not only to perform this global maximizing, but also to evaluate the worth of possible goals that the agent may adopt. The process of differentiation, earlier seen to be necessary for an actor, is now potentially under the control of the agent himself. In the exercising of agency, the choice of actions is only made possible or coherent at the point when the agent himself decides to attach certain values to possible outcomes. This prior activity becomes the basis upon which the judgments necessary for action take place. Until this evaluation has been made, action is either groundless or random for the agent has no reason for taking one decision over another.

Taylor [85] captures the very essence of the difficulty here in discussing Sartre’s example of the Frenchman deliberating over whether he should stay home and look after his elderly mother or leave and fight with the French Resistance. Because this young man must ground any decision he makes in either of two prior decisions – one that ranks filial duty over the Resistance and the other that ranks the Resistance over filial duty – his dilemma can only be resolved once he has evaluated this prior decision.

Thus, this capacity for agency permits an evaluation of the way in which alternative actions or outcomes are to be made commensurable. Before specifying the expected utility of any potential action, the agent faces the prior decision of how utility is to be assigned to outcomes. The issue, then, is not one of predicting what is going to happen, nor one of how preferences will spontaneously evolve in the future, but one of deciding which preferences are to be deemed motives for action.

A mere actor does not have this capacity for choosing among competing preference rankings of outcomes. Preferences are completely constitutive of the actor – there cannot be this further capacity for reflection that permits the agent to accept some preferences and reject others. The task facing the actor, then, is simply to choose the maximal element from his feasible set. The agent, however, must make this prior choice concerning which potential preferences he will act upon.

This distinction immediately gives rise to two additional concepts, one of which will prove useful in our investigation and one of which, I will argue, has provided much less insight than its proponents sometimes claim. Let’s begin with the less useful:

     1. Second-order preferences. This concept was first clearly articulated by Frankfurt [21]. Because agents have this capacity for choice, and are aware that they choose, they are capable of recognizing that they may, in some unspecified sense, “prefer” one potential set of preferences over a competing set. This means that our agent does not merely have desires or goals but also has desires about desires. These are exactly the second-order desires that are the conditions of personhood for Frankfurt [21:6]:

Human beings are not alone in having desires and motives, or making choices. They share these things with members of certain other species, some of which even appear to engage in deliberation and to make decisions based on prior thought. It seems to be peculiarly characteristic of humans, however, that they are able to form . . . second-order desires . . .

I am aware that this begs the question of third- or fourth- or even fifth-order desires . . . the turtles go all the way down (although it turns out we do not generally need them all). 

Frankfurt’s concept has spawned an entire literature focused generally on illuminating an issue traditionally troubling to economists: competing preferences and multiple selves. Economists, among other social scientists, noted some time ago that individuals sometimes have unstable preferences or take actions to prevent themselves from acting on their own preferences. An alcoholic may, for example, lock away his bottle during a sober moment to prevent himself from drinking later. This kind of behavior is sharply at odds with the type of rationality accorded to decision-makers by microeconomic theory. Indeed, the traditional model really has no room for decision making through time, as the rational decision-maker generally makes his consumption choice (possible contingent) once for all time and has no cause to alter his actions later. 

In reality, however, it is commonplace for multiple “selves” to compete to satisfy their own preferences. One alternates between being ready to act on the desire to flee civilization for Nepal and the desire to pursue a career in public policy. How are we to make sense of decision making in the midst of this plethora of competing preferences with only the traditional tools of microeconomics?

Second-order preferences – a mate-ranking of the first-order preference ranking of the various selves – is one analytic concept which allows us to model these competing desires. They transform the disparate mass of horizontal selves into a hierarchy of selves [19:11-3], because the most preferred preference ordering has now been identified. Indeed, second-order preferences perform the same function for competing selves in a stream of time as the static utility function did originally for our decision-maker: they provide a focal point for decision making. With competing sets of preferences (do I really desire to smoke this cigarette, or do I not?), there was no mechanism for uniquely specifying which action to take. The introduction of second-order preferences provides us with the analytic neatness that was missing with competing selves. Thus, even though there is conflict between my impish desire to dump the ice water on my bossy sister and my desire to do the proper thing and resist this temptation, it is swept away when I examine my second-order preferences – I really prefer the set of preferences that rank kindness above the momentary pleasure of soaking my sibling. Varone [89] has coined the term Topself for the preference ranking chosen by one’s second-order preference rankings. This Topself, within the utility framework, has the legitimate claim to be the self.

Despite the surface usefulness for explaining what was previously a difficulty for microeconomics, second-order preferences is not a particularly well-defined concept. Among other issues, the notion of such preferences really highlights some key problems with the concept of preferences itself – to say that all individual has both a preference to smoke and a preference to not have the preference to smoke really begs the issue of what constitutes a preference. For the most part, problems of this sort strike me as ones either of weakness of will9 or of limitations on feasible sets (i.e., with a true physical addiction, like nicotine or alcohol). There does not appear to be a pressing need to introduce layers of preferences to understand these issues. If there is indeed a “true” preference ranking arising from the Topself, then it is unclear what is added by positing these other preference rankings. They seem tofall more under the rubric of temptations – to which we may or may not succumb – and, in this case, the concept of second-order preferences may add a little analytical neatness, but it does not add much insight. As Elster [19:13] states the issue:

In one sense it is misleading to talk about metapreferences. The basis for ranking the first-order preferences must itself be a first-order preference about what the person thinks he should, all things considered. We care about preferences because they produce things we care about – actions and outcomes.

The trouble is that it is not clear which set of first-order preferences has the claim to be this Topself. There must be some distinguishing feature of the Topself that sets it apart from the cacophony of competing selves – but so far we have not managed to identify it. Elster [19:13] is typical here of pointing to how second-order preferences can be useful:

. . . once we have firmly identified the person with one of the selves, on the basis of that self’s unique capacity to form higher-order intentions about [other selves], we can use the preferences of that self to construct the meta-ranking of the person.

That is, once the Topself has been identified, an individual can and should act upon the meta-ranking of the Topself. But how to identify this Topself in the first place is never properly addressed.

An additional difficulty is that the concept of second-order preferences seems to imply a capacity to will which human beings simply do not possess. If I really could choose from among potential rankings of preferences, I would be capable of radical choice– that is, the ability to value activities through a sheer act of will. For example, I might wake up one morning and decide that I value the Kurds’ struggle in Iran over other possible preference rankings and, therefore, pack my bags for the Middle East. I do not think we have this capacity (and I will have more to say regarding this issue below).

For these reasons, the concept of second-order preferences is not sufficiently all-encompassing to address the issues for which it was developed. All we have succeeded in doing is to pile additional layers of preferences onto the decision-maker without really explaining the role for will or volition. If, within the utility framework, an individual is fully defined by his preferences, then the introduction of second-order preferences does nothing to change this. The individual is still defined by his preferences – only now they are layered upon one another.

The concept, therefore, still does not capture the crucial element of an agent’s decision-making situation: the capacity to deliberate over possible actions. If there is indeed a Topself, then we are immediately thrown back to the model of rational behavior defined by our ability to calculate to reach our goals. The agent must cut through a jungle of competing preference rankings, to be sure, but once he locates his Topself, he can heave a sigh of relief and get on with the calculating with which he is so comfortable. For this reason, while the post-Frankfurt literature that has tried to ground this concept is on the right track, it still has not investigated the key implications arising from our capacity for self reflection.

     2. Deliberative rationality. This second, and much more significant, concept captures the notion that the way in which possible states are made comparable is partially under control of the agent’s will. Indeed, Frankfurt certainly draws a clear distinction between the process of formation of second-order desires and the simple existence of the second-order desires themselves [21:7]:

. . .no animal other than man . . . appears to have the capacity for reflective self-evaluation that is manifested in the formation of second-order desires.

Second-order preferences, then, can be seen to be the outcome of the process of deliberation – they exist only after the agent has conducted the necessary self-examination and actually settled on one preference ranking rather than another – and this process culminates with Varone’s Topself, rather than the Topself itself being the primitive element. Furthermore, the agent’s evaluation is an inevitable consequence of consciousness and the agent is not exhausted by his preferences. He must be capable of the rational deliberation required to examine competing preferences. And this ability cannot be abdicated – it is absurd to envision a self-conscious agent deciding not to exercise his capacity for reflection. There are people, of course, who do not exercise this capacity – Frankfurt calls such beings wantons – but one cannot decide to be a wanton. 

A concrete evaluation example might be the trade-off between the pleasure that arises from playing golf during the afternoon versus that of going to the war history museum. How these two activities are to be ranked turns on the relative weights attached to the various sorts of pleasures that they engender. A process of deliberation must play the final role in arbitrating the issue. An agent does not simply have a preference for either one of these activities – rather, he is capable of reflecting on these activities and deciding which preference to act upon. How such reflection can proceed is the topic of Section 3.4.

The necessity of this prior decision to pin down relative weights is also recognized by Sidgwick [79:127], although he never explores the implications of this insight for his utilitarian ethic:

Shall we then say that there is a measurable quality of feeling expressed by the word “pleasure,” which is independent of its relation to volition, and strictly undefinable from its simplicity . . . This seems to be the view of some writers: but, for my own part, when I reflect on the notion of pleasure . . . the only common quality I can find in the feelings so designated seems to be that relation to desire and volition expressed by the general term “desirable”. . .

Pleasures, pains, and preferences do not arrive with ready-made ways of combining them. They do not have an intrinsic currency that is able to compare, say, the feeling I receive from listening to Bach played by the Montreal Symphony with that of a run through Central Park. It is only through the further act of will of reducing thesedistinct feelings to a common currency that the commensurability essential to the utility framework becomes possible.

Rawls [62:557] launches a similar criticism at the claims of commensurability of utilitarianism:

. . . there is the fact that there are different sorts of agreeable feelings themselves incomparable, as well as the quantitative dimensions of pleasure, intensity, and duration. How are we to balance these when they conflict? . . .The person himself must make this decision .. . . Clearly we have made no advance beyond deliberative rationality.

Both Sidgwick and Rawls are arguing that some type of deliberation must go on within the self so that disparate pleasures (and pains) are made commensurable. There must be, therefore, a self beyond that delineated solely by preferences to carry out this deliberation. If the preferences were all there were, we would be back to Hume’s dilemma of wondering what holds the bundle of perceptions together. And Nozick [53:294], in a rare display of solidarity with Rawls, also argues that the relative weights of various satisfactions is only determined through the process of making decisions:

The reasons [for decisions] do not come with previously given precisely specified weights; the decision process is not one of discovering such weights but of assigning them. The process not only weighs reasons, it (also) weights them.

The utility framework has stripped away this self capable of prior deliberation and of assigning weights to potential outcomes.

Becker [4:8], the most eloquent contemporary spokesman on behalf of the utility framework, eliminates any role for deliberation or volition in asserting not only that an individual’s preferences are fixed, but also that all individuals, at bottom, have roughly the same preferences:

Since economists generally have had little to contribute, especially in recent times, to the understanding of how preferences are formed, preferences are assumed not to change substantially over time, nor to be very different between wealthy and poor persons, or even between persons in different societies and cultures.

Becker’s assumption that individuals have similar preferences appears to spring from his statement that economists have contributed little to the understanding of how preferences are formed.10 This leap in logic can be justified empirically by presumably asserting that unobserved psychic and monetary costs close the maximizing model.

In any event, the key point is the elimination of the role of volition in determining the trade-offs among competing ends. The marginal rates of substitution between various goals have been fixed independently of the decision-maker. Interestingly, this type of explanation seems to militate against the very notion of choice Becker so vehemently defends. Indeed, a fixed relationship regulating the trade-off among possible ends has curious ties to crude behaviorism and is antithetical to the strong concept of individual choice that economists claim to support.

I think at this juncture we can now pinpoint the precise difficulty with the Principle of Sufficiency of Preferences and the utility framework. That we are agents, and not actors, implies a potential role for deliberative rationality in adjudicating among competing sets of preferences. With this insight comes the realization that the proper subject matter of inquiries into choice is how this deliberative rationality functions. The utility framework and the Principle of Sufficiency of Preferences, however, treat the subject of choice as if it concerned only the way in which individuals choose (or should choose) once preferences (either first- or second-order) have been determined. Unfortunately for adherents of the utility framework, this latter inquiry does not advance one iota our understanding of the process self-conscious agents go through to decide what potential preferences to act upon.

The obvious next step is to carefully investigate this proper subject matter – that is, the process of deliberative rationality that must underlie choice by agents. Before doing so, however, it is necessary to comment on the scope of this process. Though the concepts of agency and deliberative rationality have been introduced, and with them crucial complications in the decision maker’s problem, the considerations raised are not equally applicable to all types of decisions. In particular, the difficulty in weighing intensity versus duration of a purely carnal pleasure does not appear to present the same pressing concerns as weighing, say, duty versus honesty. In the following section, I distinguish between decision situations where deliberative rationality is integral to the quality of one’s life and those where such deliberation may not be applicable.

3.3 Simple Weighing and Agency Imagine that I am in a restaurant deliberating over whether to order the duck a 1’orange or the chateaubriand. Not a decision of any great significance and certainly not one to lose any sleep over. It does have a degree of uncertainty, of course, since one never knows exactly what the waiter will bring and the considerations raised in the previous chapter may apply. However, I do have some experience in such matters or, if I don’t, perhaps friends whose judgment I trust will advise me. Besides, a mistaken choice is not going to have any serious consequences and the extra energy I could expend on making my decision is itself costly. So, I weigh the projected satisfaction (which has been reduced to a single index) against the available options (the satisfactions of which have similarly been reduced to a single index) and make my choice. Furthermore, I may have arrived at the restaurant because I had considered the relative merits of going to a movie instead and realized that, given the differences in price, time, and satisfaction of the two activities, I would prefer going out to eat. Depending upon how much effort I expended on my decision, I also may have weighed the possibility of staying home and reading The Wall Street Journal or Schelling’s The Strategy of Conflict and then decided that either seeing a movie or eating out was preferable.

Such choices present themselves constantly – Do I have another beer while watching the football game? Do I prefer a snow blower or a shovel for the winter? Do the relative merits of the Lamborghini outweigh those of the Maserati? – and few of us have difficulty coping with them.

We make our choice by simply comparing the expected satisfactions of various alternatives. The considerations of Sidgwick and Rawls apply, of course, and one must decide how to trade off different sorts of pleasure; that is, an act of volition is required to turn the vector of possible satisfactions into the scalar of utility.

Thus, when Taylor [85:26] argues that there is nothing to the simple weighing than simply choosing, he ignores a fundamental act of volition upon which the decision-making edifice is constructed:

The simpler weigher may hesitate, as before the eclair and mille-feuilles, and his momentary preference may go back and forth. But we would not say that he envisages his situation of choice now one way, now another.

Taylor misses Sidgwick’s point regarding how distinct pleasures become commensurable. Even choices between such obviously sensual pleasures such as eclairs and mille-feuilles demand a comparison of outcomes to make choice. Thus, envisaging is possible – the eclair brings about certain satisfactions, the mille-feuille does as well. Deciding between the two means deciding on the weights to attach to the dimensions of these satisfactions. This is why Sidgwick is so insistent that every act of choice is accompanied by an act of volition (this volition, recall, is not the prime mover of action, but the prior weighting or choosing of preferences).

Nonetheless, the demand that one impose ordering does not seem terribly important in choices like this. To be blunt, it makes little or no difference whether one prefers duck or Chateaubriand, eclairs or mille-feuilles, and such choices are unlikely to call forth any careful examination of the weighting function one possesses. One simply does what one desires at that moment. The ability to scrutinize and revise one’s first preference is not exercised here and the volitional aspect recedes into the background. Taylor does hit the nail on the head when he labels such choices the “simple weighing of alternatives.”

And because these choices do not involve deeper issues of one’s life, they do not fully engage one’s capacity for agency. They can be made without serious deliberation over the relative weights of various satisfactions – and the utility framework may well be the most useful starting point.

Decisions between Chateaubriand and duck a 1’orange, however, are not the only types of decisions one faces. Inevitably, human agents face decisions in which an explanation couched in the language of the utility framework would not be sufficient.11 The notion here is that certain decisions – those that will have a definite impact on issues decision-makers perceive as being critical to their well being – demand more active evaluation in adjudicating among possible sets of preferences.

To make this clearer, I want to examine an issue that an economist might initially label “the economics of time management.” That is, how one allocates the scarce resource of time to competing and mutually exclusive activities. Time management appears particularly well-suited for the application of microeconomic theory and the underlying utility framework. Let’s also consider the decision-making problem from the viewpoint of a hypothetical faculty member, both because it is a situation with which most readers are familiar, and because it represents a relatively extreme case of the phenomenon in question.

Our faculty member arrives at his office and faces the task of deciding how he is going to spend the day. As it is summer and there are no teaching or administrative duties, he needs to allocate his research time.

The principles of Commensurability and Sufficiency of Preferences would suggest that the faculty member examine his preferences, perform the relevant calculations, and choose the research topic that yields the greatest expected utility. Even the introduction of second-order preferences would not change this much, as the faculty member must simply be careful that he acts on the preference of his Topself, and not on the preferences of a self that are not as preferred.

My claim here is that this decision sequence does not address the essential dilemma faced by the faculty member. His difficulty is that any decision concerning a research topic is necessarily parasitic on a prior decision which has chosen one preference ranking over another. He chooses to devote his energies toward a radical interpretation of Crane’s The Bridge, rather than a revamping of the classification of Celtic love stories, only because he has made a prior choice that ranks this former choice as more highly valued. If his preferences (even second-order preferences) were all there were to his identity – as the Principle of Sufficiency of Preferences implies – then this prior choice would be incoherent. Determining which research project to pursue could only consist of examining his preferences and making the correct choice.

The contrast with the duck a 1’orange case is instructive. If we asked our faculty member why he has chosen duck a 1’orange over Chateaubriand, his only answer could be that he prefers the duck. We could push the issue and ask for more specific reasons, and he might provide some – the glazing is excellent, the duck comes with rice rather than potatoes, etc. – but they may not be particularly illuminating. Preferring the duck could simply be the end of the matter.

A similar response (“I simply prefer Crane”) to a question concerning his research agenda would be nonsensical. Because he is an agent and not an actor fully defined by his preferences, choosing Crane must be grounded in some aspect of Crane chosen to be valued (assuming, of course, that his research is something important to his self-understanding). An explanation of the faculty member’s actions described in terms of choosing a maximal element would be no explanation at all, since it must presuppose the very value of the action that he himself has attached. 12

A further element that comes from this example is the necessity of articulating the judgments an agent has made. Exercising one’s capacity for agency is equivalent to being able to put into language the reasons that motivated particular judgments. It is no accident that self-consciousness and language are, at bottom, one and the same thing. Taylor characterizes this insight well [85:36]:

Much of our motivation – our desires, aspirations, evaluations – is not simply given. We give it a formulation in words or images. Indeed, by the fact that we are linguistic animals our desires and aspirations cannot but be articulated in one way or another.

A second example, moral in character, is provided by Maurice in the E.M. Forster novel of the same name.13 Maurice must adjudicate between two competing sets of preferences: one which ranks his Christian beliefs ahead of his homosexuality, in which case his homosexuality must give way, or one which ranks his homosexuality ahead of his Christian beliefs, in which case there is the implication that he can no longer remain a Christian. Introducing the concept of second-order preferences here wouldn’t get us very far, since it is precisely the ordering of the possible preference rankings that Maurice must decide. And again, articulation of this evaluation is an essential feature of Maurice’s exercise of genuine agency.

I began this chapter by pointing out that the Principle of Sufficiency of Preferences assumes that the decision-maker is completely defined by his preferences. In the last three sections, I have argued that self-conscious agents have the capacity to choose these weights, a capacity that the utility framework has been content to ignore. However, nothing has yet been said about the basis upon which such a choice can be made. Until we can say something further on how a self-conscious agent makes this choice of preferences over preferences, we have made no real progress beyond the utility framework.

Having said all this, then, the obvious next step is to ask how such deliberation can proceed so that an agent will choose one set of preferences rather than another, equally plausible, alternative. The issue is how deliberative rationality, including the relevant articulation of this deliberation, leads to the actual construction of the required second-order preferences. In the following section, I draw upon Alasdair MacIntyre’s concept of a narrative and Nozick’s description of free will to flesh out how self-conscious beings can go about making such choices.

3.4 Agency and Narratives In this section, I address the issue of how self-conscious agents can go about exercising this deliberative rationality – the true core of choice.

At the outset, three remarks must be made. First, debate over the issues raised here should be encouraged. As I asserted above, the critical question is how deliberative rationality proceeds for self-conscious agents. This section merely sketches one plausible view of this process.

Second, deliberative rationality will never be able to deliver the certainty that adherents of the utility framework find so appealing. Rationality, at least for these adherents, has always carried within it the implication that there is a uniquely rational course to follow, as we saw in Section 1.4, One may be indifferent, to be sure, but this presents no real difficulties since, if one is truly indifferent, one simply performs the action decided by the flip of a coin – heads or tails, it does not matter.14

The difficulty with deliberative rationality, however, is that there is no acceptable way to measure the rationality of an action independently of the process by which it was chosen. As there is no external measure – such as a preference ranking – all we can do is examine the process and argue over what would constitute a rational process of choosing a preference ranking. If the agent follows a previously approved process, then we have no lever to criticize his choice. Thus, for example, we may abhor the central African practice of female sexual mutilation, and be unable to understand how the women themselves can embrace a preference ranking that accepts this rite. Nevertheless, if in choosing this preference ranking they have followed all the guidelines for deliberative rationality that we have devised, we cannot criticize the practice. Thus, if we are to have any theory of rational deliberation, it cannot rule out certain “undesirable” preferences in advance but rather must argue that certain ways of justifying such preferences are illegitimate. A theory of rationality of preferences of this sort would be a broad theory of rationality, to use Elster’s [18:Chapter 1] term -although, as usual with Elster, the term is introduced and then left to be developed in some mythical future work.

Third, much of the previous thinking on this has concerned ethics, with philosophers carrying through with the exercise of deliberative rationality to support a particular ethical position. For this reason, I will start with ethics – that is, as a response to ethical utilitarianism. Nevertheless, a discussion of deliberative rationality that confines itself solely to ethics cannot possible be complete; self-conscious agents not only debate ethics, but they also make important choices that are non-ethical in nature – choices concerning career, education, relationships, etc. Any answer to the question of how decision-makers conduct this rational deliberation must, therefore, be applicable to both ethics and these other spheres of activity. This requirement will prove decisive in choosing how to describe deliberative rationality.

For the moment, then, let’s confine ourselves to ethics. Within ethics, there is a long tradition of deliberative rationality based upon the concept of an impartial judge – an individual stripped of his particular attributes – being called upon to determine the appropriate decision. In asking this judge to consider the issue independently of his particular circumstances, the hope is that we can arrive at a single, objective, rational choice. Indeed, were all individuals truly stripped of these distinguishing attributes, there would be nothing left, presumably, to differentiate their responses to moral dilemmas – everybody would necessarily give the same response. This has been the animating idea behind Kant’s categorical imperative, Adam Smith’s impartial observer, and John Stuart Mill’s theory of higher and lower pleasures.

This line of thought culminates with Rawls’ A theory of justice. This remarkable work is an extended exercise of deliberative rationality applied to the ethical dilemmas inherent in political discourse. Rawls’ thought experiment places individuals in an “original position” – thus stripped of their distinguishing attributes – and asks them to deliberate over the ethically proper social and political institutions. In this way, deliberative rationality has been applied to the design of society. By following the proper procedure to decide how to determine possible preference rankings of social institutions, a measure of legitimacy is given to Rawls’ suggestions for these institutions. The Rawls-type exercise (as well as previous ones) is a reasonable way of understanding how deliberative rationality can function, at least in the domain of ethics (though it is more difficult to see how the exercise could apply to the deliberation self-conscious agents go through in making actual choices).

Although I sympathize with the Rawlsian view, I differ on how to understand the rational deliberations of self-conscious agents when deciding what preference ranking to act upon. Individuals generally go through the process of rational deliberation with their particular attributes as one of the key inputs. Any theory of choice – whether ethical or positive – that ignores this fact winds up giving us a possibly brilliant piece of deductive reasoning that only minimally illuminates the actual conditions of choice.15 For these reasons, I assert that MacIntyre’s concept of a narrative is the best starting point for understanding how a self-conscious agent chooses to act on one set of preferences rather than another.

In After virtue, a remarkable study of the state of contemporary moral theory, MacIntyre argues that the language of morality is currently in extraordinary disarray, and that it will only resume its rightful place as an input to man’s moral thinking once much of the intellectual rubble surrounding moral philosophy is cleared away. In this section, I use MacIntyre’s concept of a narrative to explicate the nature of the problem of choice facing self-conscious agents. MacIntyre16 [42:197] asserts that our understanding of others’ actions must proceed from the placing of their actions within an intelligible story:

 . . . I [argue] that in successfully Identifying and understanding what someone else is doing we always move towards placing a particular episode in the context of a set of narrative histories, histories of both the individuals concerned and of the settings in which they act and suffer . . . It is because we all live out narratives in our lives and because we understand our own lives in terms of the narratives that we live out that the form of narrative is appropriate for understanding the actions of others.

The key point here is that as self-conscious agents we “understand our own lives in terms of the narratives that we live out . . ” In asking, then, how we can choose one set of preferences over another – how, that is, we exercise our capacity for agency – we are really asking how individuals choose to continue the narrative of which they are a part. By using these terms, the response is easier to the question of the basis for choosing second-order preferences: the self-conscious agent must be able to place his action in an intelligible narrative – his intelligible narrative.

What does it mean to have a narrative? To begin, we can sharply distinguish the concept of a narrative from any atomistic account of action. The tendency within analytic philosophy and the utility framework is to split actions (or the consumption of goods) into discrete units, which can be analyzed as free-standing objects. Thus, an individual’s decision to become a teacher rather than an investment banker is taken as prima facie evidence of a preference for teaching. Because a teacher’s monetary income is lower, it is then asserted that teaching must offer some offsetting benefits: psychic income, more leisure, etc.

The concept of a narrative is clearly at odds with this type of atomistic account. Explaining the choice to become a teacher cannot consist of simply asserting that, when all the pros and cons are appropriately weighed, the individual prefers teaching. Indeed, this response is merely another example of the utility framework arriving to explain choice after the battle has been fought – the choice of being a teacher can hardly be explained by appealing to preferences. A narrative account would begin with the notion that the selection of teaching is part of a longer story, embedded in which is this career choice. And because stories generally need to be coherent – what follows must somehow be related to what has happened previously – the process of rational deliberation performs at least part of the task imputed previously to the existence of a preference ranking in the utility framework: that of narrowing the multitude of potential choices to a manageable, and explicable, subset.

We can also see why Taylor’s insistence on articulation of preferences is critical to understand how self-conscious beings deliberate. Placing a choice within a narrative requires the ability to articulate the story. Indeed, this is precisely what we demand when we ask, for example: “Why are you going to Edmonton next week?” “Oh, because my wife’s parents are going on a three-month tour of Europe, and we wanted to take the grandchildren to see them before they went.” We could add to the narrative with further explanations – such as why seeing the grandparents is valued – but most of us have narratives that value sufficiently similar choices so that the above story is relatively coherent. An answer of “Oh, well, the tickets are non-refundable,” or worse yet, “We like Edmonton,” (with no further explanation) would be nonsensical and take us right back to the atomistic account of action at the heart of the utility framework. Frankfurt’s wanton would not be capable of this articulation.

I think that the way we generally use the word meaningful captures much of the spirit here – an action is meaningful precisely because it continues the narrative in a way that the main characters of the narrative find relatively coherent. Or, turning the issue around, the reason that there are only a relatively small number of actions that can be seriously considered at any moment is because there are only a relatively small number that can be considered meaningful. We will return to the subject of meaningfulness in Chapter 4 when the role intentions play in determining how well agents realize their objectives is more thoroughly examined.

The narrative, however, does not determine an individual’s future choices. This would be a denial of our capacity for agency. Rather, the reflection that is the hallmark of our agency can only make use of the material of one’s own experiences to generate possible paths. Particular empirical facts – including attachments, abilities, and experiences – are the basis for evaluating the worth of possible preferences. Sartre’s (reiterated in [85]) example of the young man contemplating whether to stay home with his mother or fight in the French Resistance is a case in point. The staying home option (ranking filial duty above the commitment to la patrie) is considered only because the young man has an attachment to his mother, values their relationship, and recognizes honoring that commitment as a coherent continuation of his narrative. The option to fight with the French Resistance (ranking patriotism above filial duty) is considered only because the young man is also embedded in a narrative that includes an attachment to the larger community of France, and a decision to fight is coherent in the story of this commitment.

A narrative, which does not determine future choices, is however a web of commitment that the decision-maker uses as a framework for choosing the relative weights for future decisions. When I choose to continue my narrative in one way rather than another, I commit to particular evaluations, a commitment that will generally be honored in the future. Nozick captures this point in his description of free will [53:297]:

These bestowed weights [that is, evaluations] are not so evanescent as to disappear immediately after the decision that bestows them. They set up a framework within which we make future decisions, not eternal but one we are tentatively committed to . .The situation resembles that of precedents within a legal system; an earlier decision is not simply ignored though it may be overturned for a reason . . .

That is, the individual builds the web of commitments through his prior decisions. He can reject previous decisions (though not all of them at once), but such a rejection must be based on a positive affirmation that the previous evaluation is no longer valid. That an individual could simply ignore these previous commitments would be evidence of the disintegration of individual identity. If one’s previous evaluations carried no weight in the future, then there would be nothing to hold the individual together. As Nozick [53:305] puts it: “Part of [the] nonrandom character of the weightings is shown by the life built upon them . . .” Continuity of identity, in the end, must be based on the tentative stability of these prior evaluations.

Taylor’s personal example [85] of going to a South Sea island helps illustrate the role of narrative in choice. What prevents Taylor from dropping everything he is doing, severing all his ties, and moving to a tropical island? That is, why does Taylor, as a self- conscious agent capable of deliberating over his own preferences, choose not to have the preferences that would take him to the South Seas? There are indeed many reasons to allow Micronesia’s temptations to be decisive: climate, pace of life, natural beauty, absence of pressure, less emphasis on money, and hospitable people, among others. Immediate rejection of these second-order preferences (i.e., ranking Micronesia above the daily routine) comes from the observation that they cannot be made to fit appropriately in Taylor’s narrative. It would be as if Taylor, as the playwright of his narrative, began the fifth act in a way that bore no relationship to the previous four. The conclusion here is that, just as the next scene in a play is not completely determined by the previous scenes but by the previous scenes and the direction the playwright wishes to take the story, an individual’s freedom in directing his narrative is restricted but not eliminated. Narratives get us part way toward the ideal so desired by Plato, Sidgwick, Bentham, and others – that of singling out a unique act or unique set of preferences from the myriad of possibilities. But narratives cannot get us all the way there, because there is the further act of deliberation that may yield any number of following scenes in the narrative.

I think that the term autonomy of desires, properly interpreted, is the individual characteristic that captures this aspect of deliberative rationality embedded in a narrative. One’s preferences are autonomous when one has gone through the process of rational deliberation that culminates in embracing one preference ranking rather than another. The point of this autonomy is not, as with Kant, to deny that one’s action should have anything to do with one’s particular circumstances. Rather, autonomous desires spring from considered reflection turned toward one’s particular situation and thus these particulars count. One has autonomous preferences when they are somehow one’s “own,” arising from a conscious decision on what the appropriate preferences are. One cannot start from autonomous preferences, of course (where would they have come from?), but preferences can evolve from the non-autonomous to the autonomous. Thus, having autonomous desires is the evidence of having gone through the rational deliberation of which self-conscious agents are capable. One reflects on possible preference rankings and settles upon a particular ranking because it is seen to be a coherent continuation of the individual narrative. Not moving toward autonomous preferences, and not exercising the capacity to critically examine one’s preferences, is to allow oneself to be defined by one’s preferences. Thus, an abdication of the ability to reflect takes us right back to the Principle of Sufficiency of Preferences.

Because having autonomous preferences and having conducted an exercise in rational deliberation cannot be determined from examining an individual’s preferences (recall that it is the process and not the outcome that matters), we cannot specify particular preferences as autonomous. However, there are most definitely categories of preferences that are non-autonomous. Rebellious preferences (where one desires to do something only because someone else wishes this not to be done) and those for fads (where one desires something only because others do) are both non-autonomous preferences. Elster’s concept of sour grapes [18:Chapter III], where one ceases to desire something only because it appears unattainable, is another example.

The concept of a narrative also helps bring the idea of simple weighing of alternatives into clearer focus. The reason that simple weighing is a non-issue is because such choices are generally not as relevant to the narrative. Whether one chooses the Big Mac or Whopper has little to do with how one’s life story is unfolding. The utility framework is powerful here, because it can focus on the consumption of goods and activities without asking how they fit into the larger pattern of a narrative. In short, apples and nuts are irrelevant to narratives. To move from the consideration of decisions of this type to the general problem of decision making that a self-conscious agent faces is to commit the sin of retaining the superfluous while discarding the essential.

Furthermore, the concept of a narrative provides powerful ammunition against the idea of radical choice, which is at the heart of existential philosophy. Radical choice is based on the notion that we can re-create our narratives continuously. At any moment, I can ignore my narrative up to this point and Impose a meaning on possible activities through the sheer act of will. This is the ultimate freedom that self-conscious beings possess [95]:

. . . what Sartre is saying we are free to do, is to interpret our position in the world and in time in a particular light. . . For, according to Sartre, even what we feel in a particular situation is determined by the meaning we attach to the situation .

We will examine the concept of radical choice more carefully in Chapter 4.

3.5 Conclusion In this chapter, I have argued that the simple weighing of alternatives is an inadequate description of those situations of choice that have a greater impact on one’s life. The utility framework describes choice only at the level where the simple weighing of alternatives suffices and in this way discards what is truly interesting and problematic in a decision – the dilemma that arises precisely because self-conscious agents can envisage a ranking of preferences first one way, then another. The Principle of Sufficiency of Preferences requires an individual to relinquish any claim to being capable of manipulating his own desires, even if such desires are reflective of issues that touch his life deeply.

In drawing a distinction between judgment and calculation in Chapter 2, I argued that the utility framework’s complete concern with the latter left out most of the critical issues in decision making. I make a similar claim here. Agency implies that decisions demand first the choice of preferences themselves and subsequently a calculation to choose the most preferred action. By addressing only the latter issue, the utility framework has ignored the most crucial questions regarding choice. Of course, individuals can make mistakes in calculations because rationality is finite and much could be gained from improvements in the large step from preferences to the most preferred action. This improvement, however, is secondary – most of the interesting issues of choice are embedded in the role deliberative rationality plays, and not within these questions of quality of calculation. Illuminating decision making, then, requires understanding the process by which one set of weightings is chosen over another – that is, the process of rational deliberation that self-conscious agents must go through to arrive at a set of preferences with which they can comfortably act and choose. The problem of commensurability, rather than being solved by the utility theory, has simply been restated.

Finally, MacIntyre’s concept of a narrative helps to illuminate how the self-conscious agent comes to choose a preference ranking. The requirement to meaningfully and coherently continue one’s story offers a powerful guide to choice which, though not quite decisive, provides the starting point for individual deliberation.

Returning to the issue of the contrast between individual utility maximization and profit maximization, the considerations raised in this chapter have virtually nothing to do with profit maximization. The existence of a single metric to measure success – dollars – permits the participants in a profit-maximizing enterprise to dispense with the problem of deciding what activities are to be valued. This is contingent, however, on these participants having a sufficient understanding of the overall drivers of profitability for the enterprise and perhaps on having individual incentives that encourage them to take action to maximize profits. Neither one of these conditions necessarily holds.

With respect to three derivatives of the utility framework – microeconomics, ethical utilitarianism, and decision theory – the considerations of this chapter have most often been examined in the context of ethics (as discussed in Section 3.4). Nonetheless, any positive theory of choice must address the issues raised here, and microeconomics cannot be immune to them. Indeed, positive approaches to decision making from other disciplines – sociology, or psychology, for example – must also address these same questions. This is why crude theories of behaviorism cannot be defended. Such theories are fundamentally at odds with our ability for self-reflection and choice. For the most part, decision theory has avoided the issues here by focusing its attention on a relatively narrow range of choice problems – those where the questions of preferences fall under the rubric of simple weighing or where the whole question has been assumed away. Frankly, I do not think that the tools of decision theory could be easily adapted to address the issues of this chapter. This does not invalidate the insights from decision theory – it merely implies that decision theory cannot be called upon to answer the most pressing problems of human choice.


1. Indeed, economics is often characterized as the discipline seeking to explain how individuals allocate scarce resources to competing ends. In this regard, Chapters 2 and 3 are essentially arguing that the utility framework and, hence, the economics built upon it do not illuminate these trade-offs particularly well.

Becker’s thoughts illustrate this difficulty. He argues that all individuals, regardless of society or position within the society, possess the same deep preferences. Differences in behavior are explained by appealing to different relationships between these deep and shallow preferences. Thus, the desire for a Porsche in the United States in the 1980s appeals to the same self-esteem drive that motivates the collector of skulls among some tribes in Papua New Guinea. The difficulty, unfortunately for Becker, is that we are really no closer to understanding decision making and evaluation, since we must still describe the relationship between the deep and shallow preferences. All that has been achieved is to obscure the issue sufficiently to enable economists to claim that it has been solved.

2. In the formal utility framework, the intersection of prices and preferences – with the feasible set providing a boundary condition – yields the determinate solution. If several maximal elements exist, we can arbitrarily specify one to be the decision (which one, of course, makes no difference to the decision-maker). The utility function is, of course, the mathematical shorthand for the manner in which the necessary trade-offs between competing ends are made.

3. See Endnote 30, Chapter 1.

4. Curiously, despite the surface intellectual affinity between the utility framework and individualism, there is surprisingly little room for the self here. Economics often prides itself as the social science that accentuates our ability to choose what is best for ourselves, in contrast to sociological or psychological theories that explicitly or implicitly rely upon individual irrationality.

The irony, as the passage in the text highlights, is that this is only accomplished through the elimination of any semblance of the self. The individual who chooses rationally has his decisions guided by considerations that are taken as fixed.

5. Thinking back to how economists define individuals in any of their models (e.g., let i index consumers, and $$ be consumer i’s utility function), it is easy to see that an individual is completely defined by his preferences. There is simply nothing else there.

This is all faintly reminiscent of the terrible muddle in which Hume [31] landed when he asserted that an individual is simply a bundle of perceptions. If that is all an individual is, what is the “l” that can simultaneously survey these perceptions? Hume correctly recognized that this problem was the most pressing challenge to his metaphysics.

6. Introducing the concept of volition risks burying us in 2,500 years of Western philosophy; consequently, I shall endeavor to say as little as possible here. It is worth noting, however, that by “will,” I am not referring to that obscure faculty that permits an action to proceed once it has been chosen as appropriate. As Ryle [64:63] convincingly argues, this view is an anachronism for which Descartes may be blamed:

Volitions have been postulated as special acts, or operations, ‘in the mind’ by means of which a mind gets its ideas translated into facts . . . [This story] is just an inevitable extension of the myth of the ghost in the machine. It assumes there are mental states and processes enjoying one sort of existence, and bodily states enjoying another.

When an action is performed, there are not two distinct phases: willing the action and actually doing it. There is only one – the execution of the action.

7. The Camus play, Caligula, is a case study in the damage that results from losing all basis for decision making. In Caligula’s world, judgments of states of affairs disappear and death, life, pain, and pleasure mingle in random patterns. The play is disturbing precisely because the refusal to pass judgment is closely linked to meaninglessness. Neitzsche’s Beyond good and evil is another work that goes beyond our normal process of drawing distinctions between outcomes.

8. In Kantian terms, we might say that decisions require objects of thought, and objects of thought require judgment. Indeed, Kant owes his towering (and deserved) place in philosophy precisely for this demonstration that judgment must be the fundamental faculty underlying thought. The categories are the language through which these initial judgments are expressed.

9. The topic of weakness of the will, an old one in philosophy, is treated from a number of interesting angles in J. Elster (ed.) The multiple self.

10. Becker is right about one thing here – he imposes these strong conditions because the insipid form of utility theory, which merely uses the utility function as a recording device for expressed preferences, will get us nowhere.

11.When Becker [4:8] claims that the economic approach is equally applicable to major and minor decisions, he is failing to draw this crucial distinction between those situations in which simple weighing is sufficient (and the capacity for evaluating preferences is generally not engaged) and those situations of choice that touch deeper issues in one’s life.

12. The other element which should emerge from this example is that evaluation of one’s own preferences is not exclusively moral in character. It is not always a question of comparing one’s selfish preferences with one’s social preferences, and finally admitting that, yes, one would rather act upon the social preferences.

Evaluation has to do with aesthetic or intellectual judgments as much as moral ones. Economists have generally been much too quick to reduce the issue of second-order preferences simply to this selfish versus moral choice.

13. This example is drawn from Varone [89].

14. Sen [76:68] puts the whole issue here into sharp relief:

That indifference may be a cause for dithering has often been stated. For example, Ian Little prefaced his closely reasoned attack on the concept of indifference by posing the rather thoughtful question: “How long must a person dither before he is pronounced indifferent?” But in fact there is hardly any real cause for dithering if one is really Indifferent, since the loss from choosing one alternative rather than another is exactly zero.

Burden’s ass should never starve in the economists’ model.

15. Michael Sandel [67] gives a reasonably convincing argument that is indeed the case with Rawls’ theory. I must also admit that despite my enormous personal regard for Kant’s ethical theories, he too commits the sin of building castles in the air.

16. MacIntyre certainly wasn’t the first to assert this of course – the concept can be traced back to at least Weber [95].