The scientific and policy worlds have different goals, which can lead to different standards for what constitutes "proof" of a change or phenomena, and different approaches for characterizing and conveying uncertainty and risk. These differences can compromise effective communication among scientists, policymakers, and the public, and constrain the types of socially compelling questions scientists are willing to address. In this paper, we review a set of approaches for dealing with uncertainty, and illustrate some of the errors that arise when science and policy fail to coordinate correctly. We offer a set of recommendations, including restructuring of science curricula and establishment of science-policy forums populated by leaders in both arenas, and specifically constituted to address problems of uncertainty.