Statistical Decision Criteria in the Evaluation of Information System Performance

摘要

An information system is a service facility to the management of an organization, in that decisions are made on the basis of information supplied by the system. Often the (economic) objectives of an organization can be summarized within an analysis in terms of a formal ex pression. Where the state variable outcomes occur randomly, different information systems effect different expected payoffs to the organization. As the content of information provided changes with design, the mathematical basis for computing the expected payoff conditioned on this information varies. Using the framework of statistical decision theory it is shown that different information systems correspond to applying different classes of conditional expectation operators to a randomized objective function for the organization. These formal arguments are illustrated by considering a team decision problem for the case where the payoff functional is quadratic in the action and state variables. Employing the prior-to-posterior framework of Bayes analysis, a measure of the economic performance capability of two simple information systems is shown to be a linear function of the dispersion parameters in the team a posteriori distribution on the state variables appearing in the objective function.