The ninth in a series of excerpts from Minds, Models and Milieux: Commemorating the Centennial of the Birth of Herbert Simon.
Historically, a pervasive assumption in the social sciences, in particular economics, is that humans are perfect rational agents. Having full access to information and enjoying unlimited computational resources, they maximise utility when making decisions. As is well known, Herbert A. Simon rejected this assumption, calling it a “fantasy”, for two main reasons. First, the complexity of the environment makes it impossible for humans to have full access to information. Second, a number of important restrictions impede the human cognitive system, such as limited attention and slow learning rates. Therefore, humans display only a bounded rationality and must satisfice – i.e. make decisions that are good enough, but not necessarily optimal.
Research into expertise has contributed to the question of rationality in two important ways. First, to what extent can some of the very best amongst us – super experts – approximate full rationality? Second, by what means do experts, at least in part, circumvent the constraints imposed by bounded rationality?
This chapter takes the shape of a fugue, with the themes of bounded rationality and expertise first played in the background of personal recollections, and then elaborated with a more formal survey of Simon’s research into expertise. The themes are played a third and final time with a discussion of the heuristics (rules of thumb) proposed by Simon for having a successful career in science.
Becoming an expert: A personal recollection
My collaboration with Herbert A. Simon lasted over 10 years, including 6 years spent at Carnegie Mellon. While I was working on my PhD thesis on chess players’ memory, I secured a research fellowship from the Swiss National Science Foundation to work with him. The qualifications I listed in my introductory letter to Simon were rather limited: a first degree in psychology and the title of International Chess Master. Simon, who probably saw an opportunity to reactivate research he carried out on chess expertise in the sixties and seventies (see below), but which had been dormant since, accepted to host me.
Meeting the man
I can still recall our first meeting on a beautiful morning in January 1990. His office was welcoming, but also rather disorganised, with stacks of papers and books hiding his desk. The meeting was short but cordial, and Simon gave me advice about life and housing in Pittsburgh, and briefly talked about the projects he was currently involved in.
The second meeting was my first real scientific discussion with Simon. It was actually a shock. In a polite and friendly way, Simon demolished the research line I had in mind for my PhD. The idea was to elicit a chess grandmaster’s knowledge about a small and specific domain (Rook + pawn endgames), and to build a program implementing this knowledge. The aim, inspired by research on expert systems, was to compare the amount of procedural (knowing how) and declarative knowledge (knowing what). Simon found that the project was not realistic enough (“A player like Kasparov will give you lectures on Rook endgames for several days; what are you going to do with all these data?”). In addition, he thought that the project would dovetail better with the research of his colleague John Anderson. I can still feel the panic that invaded me when he told me this, as it was an invitation to sever collaboration before the end of the first meeting! In the discussion that followed, he made it clear that he would prefer a project directly linked to the “chunking theory” he had developed with Chase in the seventies to account for chess expertise. This influential theory, and in particular the computer model MAPP (Memory-Aided Pattern Perceiver) that implemented it in part, had been severely criticised, and Simon wanted to improve on it. Thus my first lesson was that Simon, while open to other ideas, was very selective about the research lines he invested time in, and made sure that they addressed his central interests.