DECISION-MAKING PROCESSES

Leon H. Steinman

Charles K. Gordon, Jr.

Eugene B. Ross

James W. Oliver

Litton Industries

Decision-making has been characterized as a peculiarly human (a)

process. Speculatively, it has been supposed that in other species, biological or cultural factors, or a combination of them, completely determine the animal's behavior and leave it with no problem of choosing among alternative courses of action. Whatever may be true of other organisms, certainly every human being is confronted with these decision problems and, perhaps, the more complex the society, the greater the frequency and difficulty with which decisions must be made. A man in a primitive culture never encounters so difficult an issue as confronted president Truman in the summer of 1945 when the problem was: "What use, if any, should be made of the atom bomb?"

There have been many theories in the literature of economics, psychology, and sociology concerning the processes by which men make de(b) cisions. Most of these theories pertain to special cases of decisionmaking in which factors such as utility or expected utility and various kinds of estimates of probabilities are presumed to be elements of the process. There seems to have been little inquiry into decision-making processes such as that used by Huckleberry Finn, who reported his choice as follows: "I was a-trembling because I'd got to decide for ever betwixt two things, and I knowed it. I studied for a minute, sort of holding my breath, and then says to myself, "All right, then, I'll go to (c) hell."

(a) Bross, Irwin, D. J., Design for Decision. N. Y., Macmillan, 1953, ch. 1.

(b) A good survey of the literature to 1954 is provided in Ward Edwards' "The Theory of Decision Making" in the Psychological Bulletin 51: 380-417. A bibliography of 209 items is included. A more recent and more extensive bibliography is that of Paul Wasserman and Fred S. Silander, Decision-Making, An Annotated Bibliography, Ithaca, N. Y., Graduate School of Business and Public Administration, Cornell University, 1958. A third extensive bibliography, restricted to a part of the field, is contained in Games and Decisions Introduction and Critical Survey by R. Duncan Luce and Howard Raiffa, New York, Wiley, 1958.

(c) Twain, Mark, The Adventures of Huckleberry Finn, ch. 21.

Even in the limited areas of decision-making to which they pertain, the theories advanced up to the present time have had comparatively little success in predicting specific instances of human behavior. In short, very little is known concerning the actual processes involved. Perhaps the development of electronic devices to simulate these processes will help to explain the behavior of men making decisions. There is, however, a second approach to decision-making, and, correspondingly, a second use of electronic devices. This second approach, often confused with the first, provides procedures to be used in making decisions. The first approach is descriptive; it attempts to explain how choices, however "rational" or "irrational" they may be, are actually made. The second approach is prescriptive; it consists in proposals as to how decisions are to be made. It provides methods or rules for selecting one among several alternative courses of action. The techniques are proposed as yielding better solutions to problems of choice than can be expected from Huckleberry Finn's method of studying for a minute while holding one's breath.

In this second prescriptive approach, electronic devices are designed to accomplish a human function without regard to similarities between the mechanized processes and human processes; the recommendations for action which are the output of the equipment can be superior to the results of human performance because of the greater speed of operation, the greater volume and complexity of information handled, and the elimination of disturbing irrelevant considerations. It is with this prescriptive approach to decision-making-one which yields so-called "rational decisions" that this paper is concerned.

A "rational" decision-making process, whether human or electronic, may be viewed as a part of a larger system which can be represented somewhat as in Figure 1.

The system includes the processes by which the external stimuli are first obtained, the processes by which they are subsequently manipulated by what may be termed the "adaptive data processor," and, finally, the processes in the decision-making portion of our grossly oversimplified system. It should be pointed out that the model is definitely naive and, certainly, cannot possibly be interpreted as a physiological model. That is to say, in any complex physiological system, these elements need not correspond to physically separate entities at all. Nevertheless, the units represented in Figure 1 might conceivably classify major functional elements within such a system. If for no other reason than to make the ensuing presentation somewhat clearer, the model presented will serve its purpose in that it points out that the major emphasis in this paper is not on the method whereby input data is obtained, nor upon the manner in which it is organized and classified with other data. Rather,

our concern is with the output portion of the system that portion which, in manor the animal, is the step just prior to actually performing some act; namely, the decision to perform the act.

To help clarify the notion of decision-making, some terminology will be useful.

"Decisions" are called for when a "problem" presents itself and when there is imperfect knowledge as to how to act.

A "problem" exists when there is some desired state of affairs, some outcome, which it is not immediately known how to achieve. The desired state may be an internal one of the system, e. g., a homeostatic level, or an external one, one calling for a change in the environment, e. g., destroying the nearest of a set of enemy planes.

Decisions are operators (or "transfer functions") which transform a state, A, (characterized by the values of a set of internal and/or external variables) into a desired state, B (characterized by a change in the value of one or more of the A-variables).

A "goal" is a choice at a higher level, since it determines what the desired state shall be. Equivalently, a goal is a prescription to the effect that a way is to be found for transforming A into B, where A is given and B is prescribed. A goal thus delimits the pertinent or appropriate set of operators or transfer functions.

(d)

There are many different situations in which decision-making procedures are demanded. A common and useful classification groups these situations into two categories: individual decision-making, and group decision-making. Individual decision-making situations may be further classified into four types: first, the decision problem under certainty, in which it is known what the outcomes of the various alternative courses of action are; second, the decision problem under risk, in which the probabilities of the various outcomes are known, as in some gambling situations; third, the decision problem under uncertainty, in which the probabilities of the outcomes of the alternatives are completely unknown or, perhaps, not even meaningful; and, fourth, game situations, in which the outcome is uncertain because of a conflict of interest among competitors.

All these situations are ones in which a single decision is to be made once and for all. But there are a large number of decision problems of a sequential or continuous time extension nature, in which the action to be taken varies from moment to moment depending on the changing values of assorted variables. Thus, in addition to the static decision-problems of the types previously mentioned, there are dynamic decision-problems.

(d) Luce and Raiffa, op. cit., p. 13, and Chapters 13 and 14.