This work proposes a methodology for the construction of state abstraction from a set of empirically derived models of system behavior. The idea is to treat the agent in its environment as a dynamical system and augment its observation space with contextual cues extracted empirically as the agent exercises each element out of the set of available control policies--the control bias. Contextual cues are provided by the correlation between dynamic features of the agent-environment interaction and agent performance. The resulting state abstraction (observations + context information) defines also a temporal abstraction, and offers interesting answers to some of the issues pertinent to the development of hierarchical systems. Initial experiments involving an agent with impoverished sensing capabilities in a simulated, dynamic environment demonstrate that relevant contextual information can be extracted and used to enhance the agent's performance.