Book review models of bounded rationality

Posted by AJ's Blog on July 21, 2018

Book Review - Models of Bounded Rationality

Models of Bounded Rationality. Vol. I. Economic Analysis and Public Policy. Vol. 2. Behavioral Economics and Business Organization. By HERBERT A. SIMON. (Cambridge, Massachusetts & London: The MIT Press, I982. Pp. xxi +

  1. PP. xvi+5o5. /I3.Io per volume, paperback.)

The sixty papers in these two volumes are offered as ‘a reasonably complete collection of my economic works that have not been published elsewhere in book form’ (i; xviii). They are arranged in eight groups, with a brief introduction to each. The first volume, which, as the author observes, is pre- dominantly neoclassical in style, comprises Some Questions of Public Choice, Rational Choice Under Uncertainty, Technological Change, and Structure of Economic Systems; the second, which is almost as strongly behavioral, includes The Business Firm as an Organization, The Economics of Information Processing, Economics and Psychology, and Substantive and Procedural Rationality. Because of the range of Professor Simon’s work, even within the constraints of this collection, few readers would not discover unfamiliar material; but rather than attempt to survey the contents, it may be more useful to try to summarize the view of the human situation which appears to underlie them.

Whether to gain understanding or to secure a reasonable basis for action, we must attempt to simplify complexity, and any success must be incomplete and precarious. The ‘seamless web of causation’ (2; i8o) stretches far beyond our power to comprehend; therefore, although we should make the best predictions we can, we must be prepared to learn by trial and error. If experience is costly, so is the attempt to anticipate it - as Well as often being less reliable. But this is no cause to worry. ‘We must assume, as mankind has always assumed, that a reasonable allocation of our limited attention and powers of thought will solve the crucial problems facing us at least as fast as new ones arise’ (2; I 8I). The casual assertion that mankind has always assumed this is obviously false: but it is eloquent of Professor Simon’s own belief in salvation by works, and also shows him to be a pragmatist rather than a philosopher. Life is an intellectual adventure; its value lies in unresolved problems.

Efficient problem solving in a complex environment requires attention both to systems and to processes, which, in one form or another, have been the author’s professional concerns throughout his career. Although he sometimes argues that the progress of management science facilitates the centralization of decision making, he recognizes the desire of individuals within a complex system to make their own decisions, and the continuing limitations of both theory and data; thus a reasonable coupling of partial decision rules may be as much as one can hope for - and a more sensible target than the attempt to formulate some artificial system optimum. Systems analysts are usually concerned with the continuing management of complexity; they tend to seek feasible procedures rather than optimal outcomes, and ways of revising decisions as experience is gained.

As Professor Simon’s record shows, a concern with systems and processes is compatible with extensive use of neoclassical analysis - especially in mathematical form. Indeed, he is himself a notable model-builder, and clearly approves of the mathematization of social science; he takes particular pleasure in his formalization of G. C. Homan’s theory of group interaction. He even claims for unreal models the merits of ‘permitting a precise statement of the… hypothesis, and an examination of the conditions under which that hypothesis holds’ (I; 305) . Yet he insists that the language of mathematics should be used to say something about its nominal subject matter, and that neoclassical (or any other) theory should be used only where it offers a good approximation and when the relevant data are available. Thus his willingness to assume quadratic cost functions for the work with his colleagues on production scheduling and inventory control is, despite superficial appearances, not an example of Friedman’s method. The ‘false assumptions’ are justified by the feasibility of the computations which they make possible, but also, quite explicitly, as tolerable approximations within the range of decisions under consideration.

Quadratic loss functions, by contrast, which are necessary for any simple derivation of equilibrium from the postulate of rational expectations, Professor Simon regards with some suspicion (2; 438). Aggregate Cobb-Douglas production functions are inherently very dubious, and their apparent support from the data is readily explicable as a statistical artifact; they are therefore unlikely to be helpful either to understanding or to action (I; 442-54).

The scope for plausible aggregation or decomposition of complex systems is a natural problem for Professor Simon, and it is characteristic that his original paper on this subject (I; 4II-4I) should rank highly among ‘scientific publications of which I am proud’ (I; 404) - characteristic, too, that it should be a technical piece aimed at practical issues - pragmatic, not philosophical. His focus on the plausibility of either partial or aggregate models, though highly serviceable (and encouraging him to continue to believe in our ability to solve problems fast enough) does have an opportunity cost; despite recognizing the practical importance of the speed with which interactions take effect, he is, for one who prefers process to equilibrium as an organizing principle, remarkably little concerned with the significance of time in economic affairs.

Professor Simon observes (2; 383) that what an economist calls a rational decision process a psychologist might regard as habitual behavior. The economist achieves rationality by imposing bounds on the decision maker’s problem (2; 2I5). (Why, indeed, should a profit maximizer accept, without question, that no new product or technology could be discovered?) Optimization in a multi-person model requires us to assume that ‘not more than one participant is unlimitedly clever’ (2; 2I7). The interesting practical problems arise when individuals (or organizations) have to reach decisions in the face of inadequate knowledge. Whether this inadequacy is a property of the world or an unavoidable consequence of bounded rationality is, characteristically, not a question to which Professor Simon gives much attention: it is enough that the consequences are much the same.

In these circumstances, independent and simultaneous attempts at optimization are liable to be disastrous. The economic system is stabilized by its institutions (in the widest sense): the predominance of programmed behavior is a condition for originative choice. The stability of human institutions is therefore a critical issue. Professor Simon has given much attention elsewhere to the problem of organizational equilibrium; here he observes that ‘interdependence of organization units is a strong force towards conservatism’ (2; 397). Moreover, even the ‘facts’ which the decision maker accepts obtain their status as facts by a social process (2; 39I). But Professor Simon does not share the sense of fragility of such a self-supporting system which permeates the work of (for example) Kuhn, Schumpeter and Shackle.

Because of the limits of human rationality, greater information is no solution to the problem of rational decision making. Indeed, the increase of information accentuates the scarcity of attention, and therefore the need for methods of processing which will absorb information. The practical value of science is its power to replace a mass of facts with a few laws which imply them. But filtering is an active process, and Professor Simon does not perhaps pay enough attention to the possibility that socially processed theories may impose patterns on the data which may lead decision makers badly astray. This is surprising, because in effect this is the charge he brings against neoclassical economists. But for all his concern with information processing he has no theory of the growth of knowledge. He fails to use his own analysis to explain the continuing strength of neoclassicism, which puzzles him somewhat (2; 40I). When threatened with information overload, what better than to adhere to, and develop further, ‘a deductive theory that requires almost no contact with empirical data’ (2; 32I) but which is apparently rich in implications? Further- mOre, the theory is so well integrated that it triumphantly satisfies one interpretation of simplicity, offers great scope for orderly intellectual adventure, and is well protected by its own interdependencies. What perhaps needs to be added, and what Professor Simon tends to overlook - another paradox for one so appreciative of psychology - is the human desire to avoid exposure to a threatening unknown by preserving, in Adam Smith’s phrase, ‘the coherence of the ideas of the imagination’.

It is Professor Simon’s strength that he sees the unknown as an opportunity rather than a threat, since his faith is in the ability of human beings to solve problems. His ventures are not constrained to a single framework. As Alfred Marshall observed, ‘the tendency to variation is a chief cause of progress’: Professor Simon has a tendency to variation from the standard style of economics, and continues to be a chief cause of our progress.

ChangeLog

  • 2018-07-21 李亮创建