Decision Analysis: A Personal Account of How It Got Started and Evolved

I am no historian, and whenever I try, I get into trouble. I invariably forget some insignificant figures like Newton, Pascal, or Gauss, or-what is a bit more tragic-I forget some really significant figure like the fellow whose office is next to mine. What I propose to do is much more modest: I am going to concentrate on myself-on who influenced me and why, how my thinking changed over time. I was one of those returning G.I.s who resumed their college studies at the end of World War II. I studied mathematics and statistics at the University of Michigan in the late 1940s, and as a graduate student in mathematics, worked as an operations researcher on a project sponsored by the Office of Naval Research. In those days, OR was not so much a collection of mathematical techniques but an approach to complex, strategic decision making. Typically the decision entity was some branch of governmentusually one of the armed services-and they were confronted with an ill-formed problem. Part of the task of the OR expert was to crystallize the problem and structure it in such a way that systematic thinking could help the decision entity to make a wise choice. My impression is that during the war, the OR practitioners were mostly engineers and physicists with only a sprinkling of mathematicians. The problems were usually complex and required some systems thinking. Embedded in these complexities were elements of uncertainty and of game-like, competitive interactions. The most practical analyses resorted to simulations to decide which strategies were reasonable. Very often, profound insights might flow using mathematical analysis not more sophisticated than high school algebra. Indeed, there was a bias against the use of advanced mathematics. The simpler the analytical tool, the better. As the field of OR matured, mathematical researchers isolated key recurrent problems-like queueing problemsand used advanced mathematical techniques to get analytical solutions. The name of the game became simplification of reality to make mathematical analysis feasible. Then these analytically motivated abstractions were gradually made more intricate as the body of mathematical techniques grew. The trend went from elementary analysis of complex, ill-structured problems to advanced analysis of well-structured problems. In OR departments, mathematical elegance displaced the old quest for making empirically based contributions to messy real problems. A sign of maturity, I suppose. In the early days of OR, descriptions of complex realities used probabilities and one did not worry too much if these were judgmentally based, but as the field "matured," probabilities were increasingly confined to the objective domain and interpreted as long-run frequencies. There was no hint of how best to elicit subjective, judgmental information from experts about uncertainties or in identifying and structuring multiple conflicting objectives. I suspect that my continuing interest in prescriptive decision analysis had its roots firmly planted in that first academic type of job I had as a so-called "operations researcher." In that role, I was initiated into the cult that examined the world through decision-enhancing binoculars: What's the problem? Who are the decision makers? What advice would I give? A blatant, prescriptive, advice-giving orientation.