Herbert A. Simon - US grants

This project will use chess as a model system to develop and test a theory of human perception and memory, with particular attention to the memory abilities of experts and the way in which these abilities are acquired. The study of expert memory is of great importance for the improvement of teaching methods in the schools and of training and decision-making techniques in the professions and the trades. The theory will be built on the foundation of EPAM, a computer simulation of perceptual and memory processes that has, for thirty years, had remarkable success in modeling human behavior in a wide range of standard tasks, including expert memory, paired-associate and serial-anticipation learning, letter and word recognition, and concept attainment. EPAM and its extensions have explained many salient phenomena in these realms, including the serial position curve, the conditions for one- trial and incremental learning, effects of familiarity and similarity of stimuli on learning rates, expert memory for positions in chess and other games, effects of word context on speed of letter recognition, and others. This project will develop EPAM further to handle a variety of important new phenomena that have been discovered in recent years and that are not modeled accurately by the present version of EPAM. In particular, it has been found that memorization by experts is too rapid to be wholly explained by the short-term memory and chunking mechanisms in the present EPAM. On the basis of this experimental evidence and additional evidence that the project will gather, a new component of EPAM will be designed to account for these phenomena. The new EPAM will be capable of gradually acquiring in long-term memory a "template," specific to each domain of expertise, that provides a framework for rapid storage of large numbers of new observations. In the case of chess, the template will incorporate the basic structural features of the chess board and processes for storing rapidly in it information about new chess positions. While the particular templates will use chess materials, the mechanisms for building and using them will be entirely general and hence designed to account for the characteristics of expert memory in any domain. The main effort of the project will be devoted to building this extended model of perception and memory and testing it experimentally. Tracking of subjects' eye movements will play an important part in the experimental strategy. In order to make the model wholly accessible to other cognitive scientists, the system (currently written in an obsolete language) will be reprogrammed in LISP, a programming language used widely in cognitive science, so that copies of EPAM can be provided to other investigators who are studying perception and memory.