The secretary problem as a model of choice

The finite “secretory problem” in dynamic programming generates a set of mathematically optimal criteria for search strategies in sequential decisions. One of the limitations on its applicability has been its failure to allow for the uncertainty of whether or not a choice alternative which has been inspected sometime previous to the present, will be “available.” In this paper, extensions of the model are presented which incorporate not only a parameter for the uncertain availability of previous alternatives, but for the uncertain availability of an alternative currently being observed. The paper emphasizes socalled heuristic strategies which may be used to approximate optimal solutions, and which entail very reasonable demands on the processing capacity of the decision-maker. Laboratory realizations and potential applications are discussed.