The allocation of scarce spectral resources to support as many user applications as possible while maintaining reasonable quality of service is a fundamental problem in wireless communication. We argue that the problem is best formulated in terms of decision theory. We propose a scheme that takes decision-theoretic concerns (like preferences) into account and discuss the difficulties and subtleties involved in applying standard techniques from the theory of Markov Decision Processes (MDPs) in constructing an algorithm that is decision-theoretically optimal. As an example of the proposed framework, we construct such an algorithm under some simplifying assumptions. Additionally, we present analysis and simulation results that show that our algorithm meets its design goals. Finally, we investigate how far from optimal one well-known heuristic is. The main contribution of our results is in providing insight and guidance for the design of near-optimal admission-control policies.