Introduction

It is commonly believed that the rationality postulate, whatever that may mean, stands at the core of economic theory and much of social science research. Many economists go so far as to define their science in terms of the rationality assumption. I therefore try to spell out how working economists approach rationality, with the goal of explaining the practice of economic science to philosophers and other social scientists.

I do not survey the enormous literature on the methodology of economics. Much of this literature focuses on philosophy of science rationales (or lack thereof) for rationality assumptions. The writers debate instrumentalism, the use of rhetoric to discuss rationality, whether assumptions need be realistic, and whether economic propositions are or should be falsifiable, to name a few of the better-known issues. These debates have generated insight, but taken alone they give a misleading picture of what economists do. Often they focus on economics as a whole — or on one or two fields — rather than on the increasingly diverse ways in which contemporary economists conduct their research.

Philosophers, on the other hand, commonly believe that economic logic focuses on instrumental rationality, as exemplified by a Humean ends-means logic. That is, economics focuses on how to use means to achieve given ends, but it cannot judge the quality or rationality of those ends. Philosophers have put forth alternative notions of rationality, including “practical reasoning,” procedural rationality (do our mental processes for forming values make sense?

The title of this essay should naturally put knowledgeable readers in mind of a certain kind of disagreement about the nature of (rational) satisficing. Many economists, philosophers, and others have held that satisficing makes sense only in relation to a larger overall maximizing or optimizing perspective, and on such a view it is rational to seek less than the best one can only if for example one is in circumstances where maximizing is impossible or where local satisficing is a means to overall optimality. For convenience, let us call this the instrumental conception of satisficing.

It is also possible to conceive satisficing as sometimes being non-instrumentally rational, as a form of decision making that is sometimes, as we can say, inherently or intrinsically rational. This has been and still is decidedly the minority view on the rationality involved in satisficing, but the disagreement between those who maintain that all rational satisficing is instrumental and those who maintain that satisficing can sometimes be rational on non-instrumental or intrinsic grounds has been an interesting feature of the recent philosophical landscape. That interestingness may well be one reason for the existence of the present book, but I don't propose to continue this particular debate in my contribution to this volume. In fact, I know of others who will be carrying it forward here, and I very much look forward to seeing what they have to say.

What I want to do here is consider a rather different distinction having to do with satisficing.

Introduction

It is widely known that Herbert Simon introduced the notion of satisficing in order to provide an alternative to maximizing conceptions of rationality. What may be less well understood is that he sought to avoid what were (to his mind) equally unpalatable treatments of practical reasoning. Simon observed that social scientists were divided in their treatment of rationality. Economists granted homo œconomicus an absurdly omniscient rationality. On the other hand, social psychologists often explained behavior solely in terms of affect, edging rationality out of the account.

Simon's account of rationality, in contrast, aims to stake out a middle ground by taking account of the ways in which human rationality is bounded, while eschewing the elimination of rationality in favor of affect. In particular, the bounds of our rationality include limitations on our knowledge of the consequences of our actions, ignorance regarding what value we will ultimately attach to those consequences we do foresee, and finite capacity to grasp the alternatives available in any choice situation. The decision rule Simon developed within this account of bounded rationality he called satisficing, and he contrasted it with maximizing, which, he claimed, only the unbounded rationality of homo œconomicus can even hope to attain.

The intuitive motivation for a satisficing decision rule is straightforward. If the rationality of homo œconomicus is not feasible for us, and if we wish to have some explicit standard of rational choice that is feasible for us, then we need a rule that presupposes a limited, bounded rationality.

The Issue

In recent times, the view that the doctrine of maximization is too something — too demanding, too unrealistic, too stringent, or some such thing — has come into a certain vogue. Not that we are supposed to “minimize,” however: The Hegelian synthesis proposed has it that instead, the rational individual “satisfices.” Roughly, the idea is that we set a threshold such that the next sample of what we are looking for — call it F — that meets that criterion is to be chosen, even though we may be well aware that somewhere out there, there are bigger and better Fs.

The question has always been what the status of the satisficing template is by comparison with the maximizing one. Prima facie, if the rational chooser is confronted, essentially simultaneously, with two samples of F, one clearly better than the other, and he must choose between them, then he will choose the better. It seems incomprehensible that he should choose the worse, in the absence of special contexts or reasons. Is the satisficer insisting that he do so?

There is considerable temptation simply to say that one who prefers x to y even when he agrees that y is better is eo ipso irrational. If we do say this, it would be, I think, because of the practical commitments of appraisal words like ‘better’ and ‘good.’ Is to say that x is good to imply that one would choose x, other things being equal?

Introduction

The idea of satisficing either expresses the correct but relatively banal insight that the pervasive need to trade off incommensurable values puts optimizing deliberation out of reach, or else it articulates a specific, non-optimizing strategy of decision. In company with the proponents of satisficing, I believe that optimizing is seldom an apt concept for modeling deliberation. As a broad approach to human rationality emphasizing, as H. A. Simon famously did, that our limitations as deliberators mean that optimizing is rarely a rational strategy, satisficing constitutes a valuable insight; however, when satisficing is worked up into a competing strategy of decision — as it has been by a number of recent philosophers — the idea of satisficing gets into trouble, as I will show. The core idea of satisficing is that “one ceases to search for alternatives when one finds an alternative” that one deems to be “good enough.” Working this idea up into a decision-making strategy requires specifying a suitable metric of what is “good enough.” As I shall argue, showing a suitable deference to the banal facts about tradeoffs among incommensurable values while at the same time having to remain distinct from optimizing pushes the proponent of satisficing as a decision-making strategy to specify what is “good enough” in terms of a highly idealized account of what someone's preferences are.

In this chapter I examine the transparency argument for representationalism. Its central idea is this.

1. (T1) When we attempt to focus on (alleged) qualities of our experiences, we find that we cannot do so. What we find instead is only properties of the things that are represented in our experiences. The experiences themselves are transparent; we look right through them, so to speak, to the qualities of the things represented in them.

If we accept this idea at face value, then

1. (T2) Experiences must be either (a) things we are not directly aware of at all or (b) things we are directly aware of without being aware of any of their qualities.

2. (T3) Awareness of experiences without awareness of any of their qualities is both peculiar in itself and useless in accounting for phenomenal consciousness. Thus, we ought to reject alternative (b).

3. (T4) Alternative (a) is acceptable. Experiences can be regarded as brain events that represent things as being colored, flavored, and so on. In having them we are aware of the qualities of things, not of experiences. We are not directly aware of experiences; rather, we infer them as events that represent to us the qualities of things.

Therefore,

1. (T5) Experiences are transparent representers of qualities of things.

Against this argument, qualitative event realists will say that the properties that ordinary things actually have are P-properties, and nothing is represented to us in ordinary experience as having those.

The main outlines of QER can be set out in the following response to our Basic Question.

An initial motivation for realism about experiences should already be evident and can be summarized as follows. Something is happening in afterimage cases, and that something is very similar to part of what goes on in seeing. “Experiences”, and “ways in which things look (or, appear)”, are more or less well-established ways of talking about this kind of something. Differences among experiences cannot consist simply in how experiences are related to different kinds of perceived things, for there are various kinds of experiences in afterimaging, illusions, and dreams but no perceived things to which they stand in the right relations.

Many forms of materialism can be regarded as accepting a substantial portion of what qualitative event realists say about phenomenal consciousness. In rejecting QER, they attempt to give an alternative account of something that is recognized on both sides. One version of materialism, however, rejects QER in a much more radical way: it regards the qualitative events of QER as illusions. What is to be accounted for is not those events, for there are none, but only the language of seeming (looking, appearing, etc.), and the tendency of philosophers to postulate seemings (appearances, phenomenal occurrences, qualia, etc.) as presuppositions of such language. This skeptical version of materialism is closely related to the view I called “Minimalism” in Chapter 1, and its answer to the question of how color comes into experience can be given as follows.

1. (S) When Eve sees a red apple in daylight, the apple's P-red surface reflects light with some set of wavelength intensities. This light impinges on Eve's retina, where it causes changes that initiate a number of processes in the brain. Some of these brain processes may lead to behavior appropriate to red things, including reports that something red is seen. Under other conditions, some of these processes may interact with other processes and result only in “proto-judgments”, which may (or may not) lead to statements to the effect that something seems to be red but isn't. Under still other conditions, some of these processes may be substantially inhibited by other processes and thus not lead to much of anything.

2. […]

Some higher order theories model their higher order representations on perception rather than thought. For these theories, the relation between higher order and lower order states is naturally described by the equivalent terms “monitoring” and “scanning”. Theories of this kind have some distinctive features, but share with HOT theories both a two-layered structure and a corresponding interest in sensory representations that are not conscious just by themselves.

ARMSTRONG'S DRIVERS

Armstrong (1968) mentions a case that is frequently referred to in discussions of consciousness.

I shall refer to the part of such cases before one has “come to” as “autopilot cases” or as “being on autopilot”. These terms are intended as handy mnemonic labels and are stipulated to lack any theoretical content.

It would be very contentious to suppose that people on autopilot are not conscious of anything at all. I will, however, not be concerned with thoughts or perceptions that are extraneous to the driving. It will be interesting enough, and difficult enough, to consider just the perceptions of the road.

Representationalism is currently a highly favored version of materialism, and in this chapter we will consider its main attractions and its difficulties. A key part of the view can be expressed in the following summary account of how color comes into seeing.

1. (R1) Color comes into the story of a person's seeing the red apple in exactly two ways, (a) as a represented property, i.e., as a property that the person represents the apple as having (namely, red) and (b) as a property of the (surface of the) apple.

The meaning of this statement evidently depends on one's account of representation and on what one takes to be the relevant property of an apple. Representationalists have a variety of understandings on both points. I shall begin the discussion of representationalism with a generic account designed to bring out essential commitments that are common to those who claim the label. Differences among accounts will then emerge as they become relevant to assessing the merits of representationalism.

A presumed advantage of representationalism is its naturalism, that is, its commitment to recognizing no entities other than those recognized in the natural sciences. Because advances in science sometimes add new entities – e.g., electromagnetic force, black holes, or quarks – we cannot reduce naturalism to recognition of a list. Instead, we must take naturalism to accept only what is presently admitted in natural science plus whatever will come to be admitted by the methods of natural science.

In Part I of this book, I examined many versions of materialism and found them wanting. Some of them cannot be demonstrated to be false, but they fail to be good theories for other reasons, such as vacuity, unjustified skepticism, introduction of special terms that are not known to be naturalistically definable, and failure to provide advertised explanatory value. It is not good intellectual practice, scientific or otherwise, to give our assent to views that have such grievous faults.

“Materialism” suggests naturalism and respect for science. There is thus some danger that the claim that phenomenal consciousness is material will carry with it the suggestion that it has been brought safely within the orbit of scientific explanation. Since that is very far from being the case, it would be better not to suggest it. It would be better to record the existence of an explanatory gap in a terminology that keeps that fact clearly before our minds. We should say that dualism is our best theory, because we cannot plausibly deny phenomenal consciousness and we cannot give an account of it within our sciences. We know what it is to give a materialist account of life, and we have very promising ideas for further research on purely materialist theories about how our brains make us intelligent. But we do not have any comparable account of or research program about phenomenal consciousness.

Higher order, or HO, theories of consciousness claim to give a naturalistic explanation of the difference between conscious and unconscious occurrences. These theories present two kinds of challenge to QER. On the one hand, they provide positive accounts of consciousness, and a fortiori of phenomenal consciousness, that are incompatible with QER. On the other hand, HO theories also offer arguments that, if sound, would undercut the key claim that phenomenal qualities can occur only in episodes of consciousness. These two kinds of challenge are closely related and cannot be considered in complete isolation from each other. Roughly speaking, however, I will begin with the first and move on to the second in the section titled “Unconscious Pains and Tastes?”.

To introduce HO theories, let us consider cases in which people rely on assumptions without realizing it. They take in some information and they pronounce a conclusion, where the conclusion would not be rationally connected with what they took in unless a further assumption were being made. But they did not utter that assumption, either overtly or subvocally. If advised of the dependence of their conclusion on the assumption, they might deny that they relied on it and even deny its truth; and these denials might be made with every feeling of sincerity. In cases of this kind, it would be natural to say that the assumption was made unconsciously.

Are qualitative events material? Or should we hold that our conscious, qualitative events are something over and above the events that take place in the physical world? My answer to these questions will be developed over several chapters, but a rough guiding statement of it can be given: if one is willing to make “materialism” a sufficiently empty view, then it cannot be refuted, and qualitative events can be supposed to be material without self-contradiction. If, however, one wants one's philosophical views to be more than purely defensive stances, then dualism should be regarded as the more reasonable view to hold (even though it cannot be demonstrated with certainty).

To begin to clarify these rather compressed pronouncements, it will be helpful to go back to Descartes and consider the fate of an argument that is suggested by some of his formulations.

1. (D1) I am certain that I exist.

2. (D2) I am not certain that anything bodily exists.

3. (D3) I cannot be certain and uncertain of the same thing at the same time.

Therefore,

1. (D4) I am not the same thing as any bodily thing.

Perhaps the most difficult premise in this argument is (D3). Its force may be a little clearer if we consider the following reformulation of the argument.

1. (D1a) I have the property of being a thing of whose existence I am certain.

2. (D2a) No bodily thing has the property of being a thing of whose existence I am certain.

3. (D3a) For every x and y, if x and y are identical, then every property that x has is a property that y has, and conversely.