An Approach to Robust Decision Making: The Rationality of Heuristic Behavior

Empirical studies show that decisions deviate from the predictions of expected utility theory and violate the axiomatic foundations. Hence, many generalizations to non-expected utility theory have been developped. But empirically they did not provide an improvement over the standard approach. In this paper random errors are integrated in an expected utility framework. Such errors occur when agents have limited information processing capacities. A performance criterion is provided to measure the expected success of behavioral strategies. A special class of robust decision rules and its properties are analyzed. It is argued that in case of bounded rationality the evolution will select heuristics from an open class of decision rules due to performance differences.