The design of staged commitments

This dissertation develops a formulation and practical solution approaches for optimally staging commitments in a broad class of sequential decision problems involving costly review. Applications range from project review problems, wherein funding commitments are made and progress reviewed at a sequence of points in time through the life of a project, to medical observation/treatment problems, wherein a jointly optimal selection of observation frequency and treatment threshold is sought for a patient with a potentially deteriorating condition.; Though the staging of commitments is an important aspect of a broad class of decision problems, commitment staging is typically addressed informally and intuitively in the formulation phase of a decision analysis. Unfortunately, staged commitment problems are often complicated and not amenable to intuitive solution. Our approach supplements standard DA practice by providing a means for making the design of staged commitments an endogenous part of the decision problem rather than an exogenously specified aspect of the decision problem frame.; This dissertation makes three specific contributions. First, it develops a general formulation for staged commitment problems based on a Markov Decision Process framework. Second, it develops efficient algorithms for solving such problems. Third, it illustrates the applicability and significance of these ideas by developing and analyzing two large-scale example applications—one involving optimal staging of venture capital investments, and the other involving optimal management of patients afflicted with unruptured intracranial aneurysms.