Dynamic systems running in ubiquitous environments are characterized by a context that changes frequently. The adaptation of this kind of systems to the context changes is a necessary and complex task. The autonomic computing paradigm arises to tackle this complexity. In fact, autonomic systems dynamically adapt their architectures, based on a MAPE-K autonomic loop, to the context changes. Decisions during the execution of the MAPE-K loop phases must be taken in order to select the most suitable configuration of the system. In our work, we focus on decision for autonomic system adaptation. We present a state of the art which summarizes research activities dealing with decision in many fields. Thus, we present a discussion about the existing decision approaches, in order to define our approach aiming at bridging the gap of the existing decision approaches and ensuring a suitable adaptation of systems.