Algorithmic calculation of the optimality probability of decision rules (Q970494)

This paper studies the problem of choice between two alternatives, one of which is truly the best. The choice is based on preferences elicited by a group of decision makers. The skills of the individual decision makers, given by the probabilities of correctly identifying the best option, are assumed to be independently drawn from a known distribution. A method to compute the probability of a rule maximizing the probability of choosing the best option among the weighted average majority rules is presented. A special algorithm is provided for the case of the log odds of the skills being exponentially distributed.