Stuff
Random Thoughts
Tom Peters and Robert Waterman writing on Herb Simon studying chess masters
Simon, in his research on artificial intelligence finds another fascinating result that is, finally, encouraging. Looking at long-term memory, he and his colleagues studied the problem of programming computers to play chess. Within this research lies an important idea that ties together the role of the rational and the role of the intuitive. Simon started by assuming that the game of chess could be played on a strictly rationalist basis, that is, one could program the computer like a decision tree. Before moving, the computer would search ahead and examine all possible moves and countermoves. Theoretically, that can be done. However, it's not practical, for the number of possibilities is somewhat on the order of 10 to the 120th power (a trilion, by contrast, is only 10 to the 12th power). The fastest of modern-day computers can do something like that to the 20th calculations in a century. So programming our chess-playing computer to behave rationally is just not feasible.
Struck by the notion, Simon went on to research what good chess players really do. In conducting his research, he asked chess maters - the best in the world - to look briefly (for ten seconds) at games that were already in progress. The boards still containing around twonty or so pieces. He found that the chess masters could later recally the locations of virtually all the pieces. That doesn't fit with short-term memory theory at all. When class A players (one rank below masters), were asked to do the same test, they scored much less well. Maybe chess masters have better short-term memories. But here's the rub with that idea: neither the masters nor the class A players could remember where the pieces were on chessboard set-ups that were randomly generated without games in progress. Something else must be at work.
The something else, Simon believes, is that the chess masters have much more highly developed long-term chess memories, and the memories take the form of subconsciouly remembered patterns, or what Simon terms chess "vocabularies." While the class A player had a vocabulary of around 2,000 patterns, the chess master has a vocabulary of around 50,000 patterns. Chess players use decision-tree thinking, it appears, only in a very limited sense. They begin with the patterns: Have I seen this one before? In what context? What worked before?
When we start to dwell on the implications of Simon's research, we are struck by its applicability elsewhere. The mark of the true professional in any field is the rich vocabulary of patterns, developed through years of fomal education and especially through years of practical experience. The experienced doctor, the artist, the machinist, all have rich patterns vocabularies - Simon is now calling them "old friends."
Quoted in: In Search of Excellence by Thomas Peters and Robert Waterman, New York: Harper and Row, 1982, pp. 66-67, from Information Processing Models of Cognition by Herb Simon, Annual Review of Psychology vol. 30, Palo Alto: Annual Reviews, 1979, p. 363.
Local Library: In Search of Excellence (free)
Alibris: In Search of Excellence (used)
Amazon: In Search of Excellence (new)