In this paper we introduce the notion of adaptive systems, which can change their behavior during run-time. Thereby change is neither randomly nor directly pre-programmed. The change is triggered by a data processing procedure based on data acquired during the run-time of the system. In that sense the adaptation can be considered directed and unpredictable. While we have accepted, and mostly reacted, to systematic risk originating in our infrastructure, e.g. black-outs, or communication failures, we argue that adaptive systems create a new type of error-behavior that cannot be handled by classical techniques for testing systems. Here we introduce the idea of structured simulation as a means for testing the behavior of adaptive systems.