by Amartya Sen
175
t h e i d e a o f j u s t i c e r a t i o n a l d e c i s i o n s a n d
a c t u a l
c h o i c e
Are people invariably, or even typically, guided by reason, rather than, say, by passion or impulse? If the norms of rational behaviour are not followed by people in their actual behaviour, how can we seek the same answer to two rather different questions: what would be rational for a person to do? and what would the person actually do?
How can we hope, as a general rule, to tackle two quite different questions with exactly the same answer? Shouldn’t the economists who make such double use of maximization – whether through explicit reasoning or by implicit presumption – be invited to apply their minds to this?
A number of economists have indeed paid attention to systematic departures from rationality in actual choices made by people. One argument that has been invoked, in a line of reasoning proposed by Herbert Simon, goes by the name of bounded rationality.2 It concerns the possibility that people may not, in all cases, look for fully rational choices because of their inability to be sufficiently focused, or adequately steadfast or alert enough in seeking and using information that would be needed for the complete pursuit of rationality. Various empirical works have added to the evidence that the actual behaviour of people may depart from complete maximization of their goals and objectives. There is considerable evidence, powerfully presented, for example, by Kahneman, Slovik and Tversky, that people may fail to understand adequately the nature of the uncertainty that may be involved in deciding on what to expect in any specific case based on the evidence available.3
There can also be what is sometimes called the ‘weakness of will’, a subject that has received attention from many philosophers for a very long time – the ancient Greeks called it akrasia. One may know fairly well what one should do rationally, and yet fail to act in that way. People may over-eat or over-drink in a way that they themselves may think is foolish or irrational, and yet they might still fail to resist the temptations. In the economic literature, this is sometimes called ‘bounded willpower’ or ‘insufficient self-command’, 176
r a t i o n a l i t y a n d o t h e r p e o p l e and this problem too has received far-reaching attention from a number of economists – from Adam Smith in the eighteenth century to Thomas Schelling in our time.4 It is important to note that this problem is concerned with the failure of people to act in a fully rational way, but these departures in actual behaviour do not, in themselves, suggest that the idea of rationality or its demands should themselves be modified.5
The relation between rational choice and actual behaviour connects, in fact, with a long-standing divide in the discipline of economics, with some authors tending to think that it is by and large correct to assume that people’s actual behaviour would follow the dictates of rationality, while others remain deeply sceptical of that presumption.
This difference in foundational assumptions about human behaviour, and in particular the scepticism about taking actual behaviour to be identifiably rational, has not, however, prevented modern economics from using rational choice quite extensively as a predictive device.
The assumption is used often enough without any particular defence, but when some defence is given, it tends to take the form of either arguing that as a general rule this is close enough to the truth (despite some well-known divergence), or that the assumed behaviour is useful enough for the purpose at hand, which may differ from seeking the most truthful description.
Arguments for allowing certain departures from truthful description for some purpose to be distinguished from the accuracy of description, for example for their usefulness in making predictions using simple models with good track record, have been presented with much enthusiasm by Milton Friedman in particular.6 Friedman has gone on to claim that even what we regard as ‘realistic’ description should not be based on the truthfulness of the depiction, but ‘by seeing whether the theory works, which means whether it yields sufficiently accurate predictions’. This is, in fact, a very special view of descriptive realism and, not surprisingly, it has been powerfully criticized, particularly by Paul Samuelson (Samuelson has called it ‘the F-twist’). I shall not go into this debate or the issue underlying it, since they are not central to the subject matter of this book, but I have assessed the debate (and the underlying methodological issues) elsewhere.7
In scrutinizing the rationality of actual behaviour, there are also 177
t h e i d e a o f j u s t i c e some important interpretational issues that sometimes make the immediate diagnosis of irrational behaviour too rapid.8 It is, for example, possible that what appears to others as hugely irrational and even downright stupid might not actually be so inane. The diagnosis of imprudent behaviour might sometimes be based on failing to see the underlying reasons behind particular choices even when these reasons exist and are cogent enough.
Making room for irrational action is indeed important, but the diagnosis of irrationality may be a far more complex exercise than it might at first appear.9 What is important for the present work is not any presumption that people invariably act in a rational way, but rather the idea that people are not altogether alienated from the demands of rationality (even if they get things wrong from time to time, or fail to follow the dictates of reason in each and every case).
The nature of the reasoning to which people may respond is more central to this work than the exactness of people’s ability to do what reason dictates in every case without exception. People can respond to reasoning not merely in their day-to-day behaviour, but also in thinking about bigger questions, such as the nature of justice and the characteristics of an acceptable society. People’s ability to consider and respond to different types of reasoning (some of which may be well known while others not so) is often invoked in this work. The relevance of that exercise would not vanish merely because people’s actual behaviour might fail to coincide with rational choice in every case. What is more important for the purpose of the present exploration is the fact that people are, by and large, able to reason and scrutinize their own decisions and those of others. There is no unbridgeable breach there.
r a t i o n a l c h o i c e v e r s u s
s o - c a l l e d ‘ r a t i o n a l
c h o i c e
t h e o r y ’
Even when we accept, with or without reservations, the understanding that actual behaviour would not be unrelated to, or uninfluenced by, the demands of rationality, the big question of the characterization of 178
r a t i o n a l i t y a n d o t h e r p e o p l e rational choice remains. What exactly are the demands of rational choice?
One answer that has gained popularity in economics, and more recently in politics and law, is that people choose rationally if and only if they intelligently pursue their self-interest, and nothing else.
This exceedingly narrow approach to rational choice goes under the ambitious – and oddly non-denominational – name of the ‘Rational Choice Theory’ (it is called just that, somewhat amazingly, with no further qualification). Indeed, the brand-named ‘rational choice theory’, or RCT for short, characterizes rationality of choice simply as smart maximization of self-interest. It is somehow taken for granted in this approach that people would fail to be rational if they did not intelligently pursue only their own self-interest, without taking note of anything else (except to the extent that ‘the something else’
might – directly or indirectly – facilitate the promotion of their self-interest). Since human beings can easily have good reason also to pay some attention to objectives other than the single-minded pursuit of self-interest, and can see arguments in favour of taking cognizance of broader values or of normative rules of decent behaviour, RCT does reflect an extremely limited understanding of reason and rationality.
Not unexpectedly, there is a large literature on this subject, including various attempts at sophisticated defence of the understanding of rationality as s
elf-interest promotion. One of the important issues concerns the interpretation of altruistic acts based on one’s reasoning: does the existence of such a reason seen by a person indicate that he or she would actually benefit personally from acting according to that reason (‘his own reason’)? The answer must depend on the nature of the reason that is involved. If a person finds it painful to live in a society with large inequalities, and that is the reason for his trying to do something to reduce those inequalities, then his self-interest is clearly mixed up with the social goal of reducing inequality. If, on the other hand, a person wants to lessen inequality, not for diminishing one’s pain at seeing it, but because it is judged by him to be a bad thing for society (whether or not he is also pained at the sight of inequality), then the social argument must be distinguished from the personal pursuit of private gain. The different arguments involved in 179
t h e i d e a o f j u s t i c e this large – and hugely researched – subject are examined in my Rationality and Freedom (2002).10
The remarkably miniaturized view of human rationality exclusively in terms of the pursuit of self-interest will be scrutinized presently but before that I would like to consider a proposal, which I have presented elsewhere, about how rationality of choice may be – less restrictively and more cogently – characterized. Rationality of choice, in this view, is primarily a matter of basing our choices – explicitly or by implication – on reasoning that we can reflectively sustain if we subject them to critical scrutiny.* The discipline of rational choice, in this view, is foundationally connected with bringing our choices into conformity with critical investigation of the reasons for that choice. The essential demands of rational choice relate to subjecting one’s choices – of actions as well as objectives, values and priorities – to reasoned scrutiny.
This approach is based on the idea of a link between what would be rational for us to choose and what we have reason to choose.
Having reason to do something is not just a matter of an unscrutinized conviction – a strong ‘gut feeling’ – that we have ‘excellent grounds’
for doing what we choose to do. Rather, it demands that we investigate the reasons underlying the choice and consider whether the alleged reasons survive searching and critical examination, which one can undertake if and when the importance of such self-scrutiny is understood. The grounds of choice have to survive investigation based on close reasoning (with adequate reflection and, when necessary, dialogue with others), taking note of more information if and when it is relevant and accessible. We can not only assess our decisions, given our objectives and values; we can also scrutinize the critical sustainability of these objectives and values themselves.11
* While some technical issues, including some mathematical ones, are involved in developing this point of view, the main argument can be fairly easily understood in terms of seeing rationality as conformity with reasons that one can sustain, even after scrutiny, and not just at first sight. For a general presentation and defence of this approach, see my essay ‘Introduction: Rationality and Freedom’ in Rationality and Freedom (2002). The more technical questions are addressed in essays 3–7 in the same volume. See also Richard Tuck, Free Riding (Cambridge, MA: Harvard University Press, 2008).
180
r a t i o n a l i t y a n d o t h e r p e o p l e This is not, of course, the same thing as demanding that every time we choose something, we must undertake an extensive critical scrutiny
– life would be intolerable if rational behaviour were to demand that.
But it can be argued that a choice would count as rational only if it would be sustainable had a reasoned critical scrutiny been undertaken.
When the reasons for a particular choice are established in our mind through experience or habit formation, we may often choose reasonably enough without sweating over the rationality of every decision.
There is nothing particularly contrary in such norms of sensible behaviour (even though we may sometimes get deceived by old fixed habits to which we are wedded when new circumstances demand a departure). A person whose habitual choice of after-dinner coffee tends to be decaffeinated, even though she enjoys decaffeinated coffee less than regular coffee, may not be acting irrationally, despite not undertaking a reasoned scrutiny on each occasion. Her habit may be based on her implicit reasoning because of her general understanding that coffee at that hour would keep her awake, as shown by her past experience. She does not have to recollect on every occasion the tossing and turning in the bed that would result from her drinking regular coffee at that hour. Sustainable reasoning can exist without undertaking explicit scrutiny on every occasion.
This general approach to rational choice – seeing rational choice as choice based on sustainable reasons – has appeared to some to be so general that there has been an evident temptation to think that it could not possibly amount to saying anything much at all. In fact, however, the understanding of rational choice as choice based on sustainable reasoning makes quite strong claims of its own, while rejecting a variety of other claims about the nature of ‘rational choice’.
Indeed, seeing rational choice as critically scrutinized choice is both exacting and permissive.
It is exacting in that no simple formula (such as the maximization of self-interest) is automatically taken to be rational without subjecting that formula to a searching scrutiny, including critical examination of both the objects to pursue and the constraints of sensible behaviour that one may have reason to follow. For example, the narrow view of rationality reflected in the so-called rational choice theory would not have any immediate claim to being taken to be appropriate.
181
t h e i d e a o f j u s t i c e It is worth noting here that the general framework of maximization which provides a mathematical structure to disciplined choice is itself much broader than maximization of self-interest in particular.* If one’s goals go beyond one’s own specific interest and bring in broader values that one has reason to advance or appreciate, then the maximization of goal fulfilment can depart from the specific demands of self-interest maximization. Further, if one has reason to accept certain self-imposed constraints of ‘decent behaviour’ (varying from following safety rules of orderly exit without jostling one’s way to the exit door as the fire alarm sounds, to more mundane practices like not racing to take the most comfortable chair in a social gathering, leaving others far behind), then goal maximization subject to those self-imposed constraints can conform to the broad demands of rationality.†
If the approach of rational choice as ‘critically scrutinized choice’
is, in this sense, more exacting than following the simple formula of self-interest maximization, it is also more permissive in that it does not rule out the possibility that more than one particular identification
* The ability of the mathematics of maximization to take on board different types of constraints and varieties of goals (including menu-dependent preferences) is discussed in my ‘Maximization and the Act of Choice’, Econometrica, 65 (1997). See also Rationality and Freedom (2002). I should, however, note here that the analytical characterization of maximization does not quite capture the way in which that term is often used in loose speech. If I am told that I should be aware of the fact that ‘Paul is a ferocious maximizer’, I would not be tempted to think that Paul relentlessly pursues the maximization of the social good in a selfless way. The common and distinctly unflattering use of the term ‘maximizer’ is perfectly fine within its own linguistic context, but it has to be distinguished from the analytical characterization of maximization.
† Sometimes these rules of decent behaviour are also in one’s self-interest in the long run, but the justification for it need not be based only on the pursuit of self-interest.
The point to ascertain is not so much whether a practice serves one’s self-interest (that can be one important reason among others to follow a rule), but whether one has reason enough to follow that rule (whether based on self-interest
or on some other reasoned ground). The distinction is analysed in my ‘Maximization and the Act of Choice’, Econometrica, 65 (1997). See also Walter Bossert and Kotaro Suzumura,
‘Rational Choice on General Domains’, in Kaushik Basu and Ravi Kanbur (eds), Arguments for a Better World: Essays in Honor of Amartya Sen, Vol. 1 (Oxford: Oxford University Press, 2009), and Shatakshee Dhongde and Prasanta K. Pattanaik,
‘Preference, Choice and Rationality: Amartya Sen’s Critique of the Theory of Rational Choice in Economics’, in Christopher W. Morris (ed.), Amartya Sen, Contemporary Philosophy in Focus series (Cambridge: Cambridge University Press, 2009).
182
r a t i o n a l i t y a n d o t h e r p e o p l e of what can be chosen with reason could survive a person’s critical scrutiny. One person could be rather more altruistic than another without either of them violating the norms of rationality. We might also find one person more reasonable than another, invoking – perhaps implicitly – what our idea of being ‘reasonable’ is in a social context (as John Rawls has done), but that would not make the latter person necessarily irrational. Though the demands of critical self-scrutiny are exacting, it can still allow a variety of competing reasons to receive contending attention.*
There is, however, a rather straightforward implication of this permissibility that is worth commenting on. Since the demands of rational choice might not invariably yield the identification of a unique alternative that must be chosen, the use of rational choice for predictive purposes cannot but be problematic. How can rational choice point to what would be actually chosen if there is more than one alternative that could count as rational? It is one thing to accept the need to understand the nature of rational choice because of its own importance and also for its relevance in analysing actual choice, but it is quite another to expect that an understanding of rationality of choice could be immediately translated into the prediction of actual choice based on the set of choices that all count as rational, even when human beings are assumed to stick invariably to choices that are rational.