This study investigates the effect of reported probabilities on decision making performance in a managerial task. After discussing the importance of such research, the cognitive complexity approach is identified as the appropriate theoretical approach to this study of human 'information' processing.
Both the normative and descriptive decision making literature in psychology and accounting is reviewed with regard to the concepts of data, information, uncertainty, data processing, and reported probabilities: and also with respect to four tenets of man's decision making behaviour. From this, a general theoretical framework of decision making, using concepts from both normative and descriptive research, is proposed. The Schroder, Driver and Struefert (SDS) literature is then reviewed within the perspective of the general framework, and seven laws of interaction involving decision making performance are developed.
The five variables identified in the literature as potentially important to performance are- incorporated into a factorial research design. Empirical indicators are then developed for these variables to permit seven research hypotheses to be stated. To test these hypotheses a laboratory experiment was conducted which required 134 non-volunteer university students to make three corporate strategy decisions. Their confidence in these decisions, the time taken, and their accuracy were then analysed using three separate ANOAs.
The results generally support the hypotheses. They indicate that when the probabilities of two mutually exclusive events are close, data load is at its highest. The use of a decision model reduces or eliminates the effect of data load on decision making performance. The relationship of time to data load may be confounded by the adoption of time saving heuristics by experienced highly intelligent and motivated decision makers. Last, at high data loads decision making confidence is overrated, but does decline, as does decision making accuracy, over data load.
These results suggest that the data load produced by an accounting information system is an important consideration in the design of the system. This importance is particularly evident in electronic data processing systems which may frustrate the assessment of data credibility, and summarise in a way that interferes with human decision making.