One purpose of an accounting system, or any information system, is to provide a set of signals designed to communicate descriptions of certain past phenomena believed to be decision relevant in the future.' However, it is difficult to evaluate a given or proposed information system because of its complexity. This complexity arises in part because of the interrelated problems of which phenomena to describe and how best to describe them. Since complexity hinders information system evaluation, decomposition of the total interrelated system into a number of less complex subsystems for evaluation purposes is often desirable. Unfortunately, most decomposition will introduce errors into the analysis. But if we can predict the resultant errors, decomposition can still provide a useful surrogate evaluation method. The purpose of this paper is to explore the decomposition approach to information system evaluation, at the conceptual level, relying heavily on Feltham's recently proposed model for predicting the value of information in the single decision case.2

Performance evaluation models are usually constructed and implemented without giving formal consideration to desirable adaptation of a decision during its actual implementation. Moreover, it is generally acknowledged that such adaptation may be desirable and is usually accomplished in some heuristic fashion. Since the performance evaluation model, through both information flow and behavioral effects, may affect this process, the possibility of altering the evaluation model in an advantageous manner becomes an issue. Control of the decision implementation interface is a poorly structured, little understood, problem in management science that is of direct concern to the accountant. Using a firm simulation experiment, we have examined the differing implementation effects of three alternative performance measurement models. We define an implemented decision as a specific act that an organizational decision unit attempts to achieve, such as an output of x units during some specified time period. The final result of this implemented decision is not necessarily equal to the priori-desired result; and the relationship between the implemented and ultimately achieved act we term the decision implementation interface. More important, however, the results are sufficiently intriguing to warrant replication of the experiment.

The article discusses the nature of decision models and information evaluation models. The first section presents a general framework, or model, for information evaluation. The second examines the nature of the operational model that would be constructed in order to apply the general framework in a specific setting. The information evaluation process is viewed in cost-benefit terms; and the cost-benefit, or value, calculation is developed from the point of view of an individual who decides what information to supply. Different signals may result in different decisions, and, therefore, the information evaluator must develop a conditional probability distribution over the future events for each possible signal. Decision models seldom represent as controllable variables all actions that must be taken. Most mathematical models do not attempt to include detailed predictions of individual events. One simplified and well-known method for selecting information systems is to base the selection on personal opinion. Some insights that can be developed by such an exploration were then related to information choice issues and information research methods.