PSYCHOLOGICAL SCIENCE General Article THE REWARDS AND HAZARDS OF COMPUTER SIMULATIONS

Although computer simulations and other modeling tools hm'e assllmed a pivotal role in cognitive psychology, their IItility continues to be questioned by some psychologists. This article presents several examples that illllstrate both rewards and po-tential hazards associated with the simlllation approach. Sim-Illations can provide the formal framework necessary to disam-biguate new ideas, they can explore the implications ofcomplex models, and they can predict seemingly cOllllterilltuiti,'e find-ings or uncover hidden relationships. At the same time, care mllst be taken to avoid pitfalls that may arise when compllter code inadvertently differs from the intended specifications ofa theory, or when predictions derive not fromfimdamelltal prop-erties of a theory but from pragmatic choices made by the modeler. The simulation of human behavior by computers has become a burgeoning enterprise: As early as 1971, Star-buck and Dutton compiled a bibliography encompassing 1,921 articles concerned with computer simulations. That number appears particularly staggering because it dates from a period in which only a handful of psychologists were conversant with simulation techniques. During the two succeeding decades, the number of psychologists conducting simulations has increased considerably, and computer simulations are now nearly as common as ex-periments in some parts of the literature. However, this trend has not been without opposition, and demands have recently been made for the complete elimination of mem-ory models or theoretical constructs of any kind (Wat-kins, 1990). That position, in turn, has not gone unchal-lenged and has been countered by Hintzman (1991), who illustrated the utility of models by pointing mainly to known limitations of human reasoning, and how those could be circumvented by the proper use of models. Although stimulated by recent controversy, this article was not written to continue the debate between modelers and behavioral empiricists. Instead, the article illustrates the rewards offered by computer simulations, as well a