Herbert A. Simon - Artificial Intelligence and Psychology

Artificial Intelligence and Psychology

Simon was a pioneer in the field of artificial intelligence, creating with Allen Newell the Logic Theory Machine (1956) and the General Problem Solver (GPS) (1957) programs. GPS was possibly the first method of separating problem solving strategy from information about particular problems. Both programs were developed using the Information Processing Language (IPL) (1956) developed by Newell, Cliff Shaw and Simon. Donald Knuth mentions the development of list processing in IPL with the linked list originally called "NSS memory" for its inventors. In 1957, Simon predicted that computer chess would surpass human chess abilities within "10 years" when, in reality, that transition took about 40 years.

In the early 1960s Simon wrote a paper responding to a claim by the psychologist Ulric Neisser that machines might be able to replicate 'cold cognition', e.g. processes like reasoning, planning, perceiving, and deciding, but could not replicate 'hot cognition', including desiring, feeling pain or pleasure, and having emotions. Simon's paper was eventually published in 1967. It was ignored by the AI research community for some years, but later became very influential e.g. indirectly through the work of Sloman and Picard on emotions.

Simon also collaborated with James G. March on several works in organization theory.

With Allen Newell, Simon developed a theory for the simulation of human problem solving behavior using production rules. The study of human problem solving required new kinds of human measurements and, with Anders Ericsson, Simon developed the experimental technique of verbal protocol analysis. Simon was interested in the role of knowledge in expertise. He said that to become an expert required about 10 years of experience and he and colleagues estimated that expertise was the result of learning roughly 50,000 chunks of information. A chess expert was said to have learned about 50,000 chunks or chess position patterns.

Simon was also interested in how humans learn and, with Edward Feigenbaum, he developed the EPAM (Elementary Perceiver and Memorizer) theory, one of the first theories of learning to be implemented as a computer program. EPAM was able to explain a large number of phenomena in the field of verbal learning. Later versions of the model were applied to concept formation and the acquisition of expertise. With Fernand Gobet, he has expanded the EPAM theory into the CHREST computational model. The theory explains how simple chunks of information form the building blocks of schemata, which are more complex structures. CHREST has been predominantly used to simulate aspects of chess expertise.

He was awarded ACM's A.M. Turing Award along with Allen Newell in 1975. "In joint scientific efforts extending over twenty years, initially in collaboration with J. C. (Cliff) Shaw at the RAND Corporation, and subsequentially with numerous faculty and student colleagues at Carnegie Mellon University, they have made basic contributions to artificial intelligence, the psychology of human cognition, and list processing."