Various medical groups are continually putting out advice as to health-inducing behavior (diet, exercise, and so forth). Individuals who follow this … - Kenneth Arrow

Among economic phenomena which have in some way been tied up with the existence of uncertainty, three classes may be distinguished: (1) those which by their very definition are concerned with uncertainty; (2) those which are not related to uncertainty by definition but nevertheless have no other conceivable explanation; (3) those whose relation to uncertainty is more remote and disputable.

The businessman may be compared with two other types of individuals who are essentially concerned with behavior under uncertainty ― the scientist and the statistician. The scientist must choose, on the basis of limited information, among the innumerable logically conceivable laws of nature, a limited number. He cannot know whether his decisions are right or wrong, and, indeed, it is none too clear what is meant by those terms. There is a long history of attempts to reduce scientific method to system, including many which introduce probability theory, but it cannot be said that any great formal success has attended these efforts. If we were to compare the businessman to the scientist, we would be forced to the melancholy conclusion that little of a systematic nature can be said about the former’s decision-making processes.
The statistician typically finds himself in situations more similar to that of the businessman. The problem of statistics can be formulated roughly as follows. It is known that one out of a number of hypotheses about a given situation is true. The statistician has the choice of one of a number of different experiments (a series of experiments can be regarded as a single experiment, so that drawing a sample of any size can be included in this schema), the outcome of any one of which is a random variable with a probability distribution depending on which of the unknown hypotheses is correct. On the basis of that outcome, the statistician must take some action (accept or reject a hypothesis, estimate the mean of a distribution to be some particular value, accept or reject a lot of goods, recommend a change in production methods, and so on), the consequences of which depend on the action taken and on the hypothesis that is actually true.

Ever since I encountered Hicks’s Value and Capital while I was still a graduate student, I had the aim of completing and extending his vision of the economic system in its purest form. This was not because I believed that the economic world was perfectly competitive or that it was clearly self-equilibrating; after all, Chamberlin, Robinson, and Keynes were dominant intellectual influences, and I had the even more powerful influence of the facts of massive unemployment and large corporations. But the idea that the economic world was a general system, with all parts interdependent, seemed (and seems) to me to be an essential of good analysis. I regret what appears to be a revival of single-market thinking both among monetarists and among some of the younger empirical analysts. Then as now, the only game in town that offered a general system of economic interdependence was general competitive equilibrium, an idea to which the name of Leon Walras is imperishably linked. At least, such a system would provide a starting point for analysis of the market’s imperfections.