Will as international bargaining: implications for rationality.

• Author: Ainslie, George
• Position: Preferences and Rational Choice: New Perspectives and Legal Implications

Rationality has been understood as conducting yourself according to reason rather than passion. In modern times this endeavor has become synonymous with maximizing your expected goods, with the value of expected but delayed goods discounted exponentially. However, behavioral research has found a robust tendency for delayed goods to be discounted hyperbolically, that is, for their value to be divided by their delay. This finding supplies a simple hypothesis about the origin of irrationality, but greatly complicates the problem of rationality, since it depicts a limited warfare relationship among interests within an individual. Recent research on the combining properties of hyperbolically discounted rewards supports the hypothesis that a person's will arises from a prisoner's-dilemma-like relationship among successive motivational states. This picoeconomic hypothesis provides a mechanism for both the strength and "freedom" of the will, and predicts pathologies of overcontrol that make strength of will something very different from pure rationality. This approach offers insights into current puzzles about criminal responsibility and the disease model of addiction.

INTRODUCTION

Rationality is an ancient concept, one that Plato contrasted with passion to form a dichotomy of choice principles. (1) Through the ages rationality has meant the good way to make choices, the way that will maximize your satisfaction with the outcome. As such, it has been a norm rather than a description of actual behavior. However, since utility theory has postulated that people always maximize their expected utility, rationality has acquired a descriptive implication: the rational is what anyone inevitably will do whenever she is aware of the true contingencies she faces. Irrationality then is merely error, the product of some fallacious valuation process. Modern rational choice theory (RCT) thus aims not only at normative optimality but also at factual accuracy; it consists of "a series of assumptions about how people respond to incentives." (2) Yet it inevitably retains a normative implication as well, the implied contrast with the phenomenon of irrationality.

As a descriptive theory, RCT has come under attack from two disparate directions. People wary of reductionist science complain that it under-recognizes empathic transactions; they claim that it depicts as natural--and thus promulgates--a selfish, money-grubbing society. (3) At the same time, empirical researchers find that it fails to predict important examples of behavior exhibited by well-informed subjects in experiments within behavioral science. (4) Furthermore, the examples documented by systematic analysis are only a small proportion of the behavior patterns that people say they do not want but seem unable to give up. Seemingly free choice has led not only to alcoholism and drug abuse in a significant minority of people, but also to an epidemic of overeating, credit card abuse, overconsumption of passive entertainment, and other bad habits too widespread to be diagnosed as pathological. (5)

RCT arose not so much from empirical research as from a theoretical analysis of what decision strategies will dominate in marketplaces. (6) To prevail over any significant period of time, an intention must be stable, and stability requires the standard properties of rationality--particularly commensurability, transitivity, and invariance across contexts. (7) However, this approach makes rational choice theory a set of rules for winning play--a normative model--rather than a description of how choice actually works. Even in the far simpler world of game theory, human subjects in experiments notoriously fail to follow obvious strategies that would increase their success. (8)

Psychologist Daniel Kahneman and his collaborators have directly studied human utility maximization. They estimated "objective happiness" by taking subjects' numeric self-reports of happiness moment by moment and calculating the area under the resulting curve over time. (9) According to the researchers, "[l]ogical analysis suggests" that the utility that a person derives from a particular event should be equal to the integral of all the instants that make up that event. (10) The integral that the researchers derived from this experiment did not reflect the way that the same subjects chose between the very experiences that they had evaluated this way. Subjects did not prefer those experiences with the greatest summed happiness (or least summed unhappiness), but displayed various perceptual distortions, particularly overvaluation of the greatest momentary reading and the latest reading. (11) These experiments demonstrate that real world utilities, as evaluated by RCT, fail tests for transitivity and invariance and, therefore, that conventional RCT is flawed as a descriptive theory.

In this Article, we present a utility-based model that fixes the major problems of RCT. We argue that decisions are determined in a single intrapersonal marketplace on the basis of a unitary selective principle--reward (Part I)--but the basic shape of the discount curve for this reward creates conflicting interests within the individual (Part II). These conflicts can be partially resolved by perception of a prisoner's-dilemma-like relationship among successive motivational states, which generates will (Part III), but which does not approach rationality closely (Part IV). This model has practical implications for law and economics, the direction of which we can only suggest (Part V).

1. ALL REWARD MUST BE COMMENSURABLE

   While RCT is unable to account for certain behavioral phenomena, in rejecting it there is a risk of throwing out the baby with the bath water. It is true that people do not behave so as to maximize attainment of a stable and ordered set of goals. Indeed, we will argue that they do not maximize any quantum without regard to the time at which they make their choice. Such a finding, however, does not require the conclusion--often encountered in cognitive psychology--that choice is irreducibly particularistic and not constrained to maximize anything. We do not have to abandon RCT's assumption that motivated behavior occurs within a single internal market, with a single currency of transaction. Indeed, research in both neurophysiology and behavioral psychology points to just such a market.

   Over the last half-century neurophysiologists have located, with increasing precision, brain sites that control the selection of behaviors. From early on, it was known that animals would work to receive electrical stimulation in the medial forebrain bundle and nucleus accumbens. (12) Rats that can press a lever to get stimulation in these areas have been observed to press continuously for hours, until they become too weak to go on. (13) The same portions of their brains, which form part of the mesolimbic (midbrain) reward circuitry, were later shown to be those excited by cocaine (14) and by all other rewarding drugs that have been studied. (15) Furthermore, advances in brain-imaging technology have made it possible to observe the reward process in normal human volunteers. These studies have shown that activity in mesolimbic reward circuitry accompanies even minor rewards such as winning a dollar, (16) receiving a squirt of pleasant tasting juice, (17) or viewing an attractive face. (18) Thus, there is an observable physical process that responds proportionately to known rewards.

   It may turn out that localization of reward is more complex than current data suggest. However, commensurability of incentives is, in the end, a logical requirement of any theory of voluntary behavior; otherwise there would exist choices that could not be made. (19) As neurophysiologists Peter Shizgal and Kent Conover point out, the ultimate basis of choice must include a comprehensive marketplace of incentives. (20) They state that "[f]or orderly choice to be possible, the utility of all competing resources must be represented on a single, common dimension." (21) A model in which the activity of a quantitative selective mechanism determines value and hence choice need not require the maximization of total utility, but prevailing choices must be doing something over time that induces more of this selective process (hereinafter reward) than their rejected alternatives did.

   The data from neurophysiology only add anatomic specificity to the vast literature on behavioral psychology, which has shown that choice is exquisitely sensitive to small changes in incentive. (22) Unlike neurophysiology, behavioral psychology has extensively studied the effect of delay on the incentive value of reward. This study of delay has suggested a way to reconcile subjects' failure to maximize expected reward with the strict determination of choice by reward.

2. THE VALUE OF REWARD IS INVERSELY PROPORTIONAL TO DELAY

   It has long been known that people discount the value of delayed goods, although some writers from Plato (23) to the Victorian economist Jevons (24) have called such discounting irrational. However, no one now thinks that rationality means maximization of reward without regard to timing--that $1001 a year from now is worth more than $1000 now. Thus, the concept of reward maximization has had to include some kind of discounting from the moment of expected delivery back to the moment of decision. Financial markets long ago established a norm for how this discounting should take place--the loss of a constant proportion of remaining value per unit of time, or exponential curve, which is the only function that will not lead to changes of relative valuation among goods at different delays as time passes. People adopted this curve to such an extent that, as utility theory took mathematical shape, this curve was assumed to depict not only the normatively rational discount rate but also the one that people follow spontaneously.

   However, precise preference experiments in the last...