Extreme events are significant determinants of the character and evolution of many natural and human systems. When extreme events occur at the interface between natural and human systems, they are often called "disasters." Here, we use a systemic, contextual view of disasters to construct a framework for organizing research and policy. Within this framework, reduction of vulnerability is the organizing principle, and decision processes (which lead to reduced vulnerability) are the fundamental unit of analysis and action. Scientific research is connected to decision processes through knowledge-integrating activities such as prediction, observation, and heuristics. But the value of research depends on its capacity to enhance decision-making capabilities. Our goal is to define an approach by which policy-relevant research questions can be more readily recognized. and societally valuable (i.e., vulnerability-reducing) knowledge can be more effectively created and used.