Making decisions under the influence of memory.

Abstract

John FoxUniversity of Sheffield, Sheffield, EnglandTheories of decision making are reviewed, particularly those that shed light onclinical decision making. Two approaches are examined in detail, traditionalprescriptive theories (derived from probability theory) and theories based onnonprobabilistic inference processes with a supportive memory mechanism. Anexperiment using a laboratory analogue of clinical diagnosis supplied data aboutseveral aspects of the decision process including information seeking and theterminal decision. A second task interleaved with the diagnostic task was de-signed to provide some insight into the subjects' organization of the diagnostictask material in memory using the proposition confirmation/disconnrmationmethod. Systematic effects of task structure on the availability of knowledgeabout the task were obtained. In two computer simulations the traditionalprescriptive approach to decision making gave a moderately good account ofthe behavior observed in the diagnostic task, but the cognitive theory based onnonprobabilistic inference rules interacting with simple recall processes gaveequal or better quantitative predictions. It is concluded that the nonprobabil-istic approach holds considerable promise for the description of complex deci-sion processes.From an information-processing standpoint,human decision making depends on the inter-action of two main cognitive components: thevarious logical and inferential processes thatcontribute to the final decision and, at leastconceptually distinct, the processes that makethe knowledge or data available that are re-quired for the decision task at hand. Untilrecently, however, decision theorists havebeen more preoccupied with the mathematicaldescription of human decision behavior thanwith a general cognitive account of its processes(see, e.g., Slovic, Fischoff, & Lichtenstein,1977). The tide has recently begun to turn withwork like that on heuristics by Tversky andKahneman (1974), but we have not yet reachedthe point where decision processes can beMany thanks to Richard M. Young for his criticalreading of an early draft of this article, and particularthanks to Vicki Bruce, whose careful and insightfulcomments contributed so much to the final manuscript.Requests for reprints should be sent to John Fox,Medical Research Council, Social and Applied Psy-chology Unit, University of Sheffield, Sheffield S10 2TN,England.described in any detail, certainly not decisionsof natural or everyday complexity.This study looks at a more naturalistic de-cision task than is usual and examines somealternative theoretical frameworks for charac-terizing the decision process. The contrast thatemerges is that between the decision theoristswho have tried to develop algebraic and sta-tistical models that are largely fre of performa-tive elements and the thrust elsewhere incognitive psychology, which has been towardmore complete accounts. The decision theoristshave, notably, failed to suggest what sorts ofinference or problem-solving processes de-cision makers typically engage in or howknowledge influences and interacts with thoseprocesses.For illustration we focus on a task modeledon clinical decision making, a field in which theimportance of knowledge is particularly promi-nent. Several possible theoretical accounts ofclinical decision behavior are briefly reviewed,and a decision study is described, which isdesigned to shed light on some of these pro-posals. Finally, two computer simulations of