Discrete-Event Dynamic Systems: A DEDS framework models the behavior of a continuous process in terms of discrete events observed at run-time. Axioms in a DEDS specification can be used to prove that certain discrete events can't occur by limiting the set of admissible control actions. Using these tools, control decisions can be focused on safe and relevant actions while learning to construct policies for behavior. This is one of the ways that autonomous learning machines can be directed to acquire important control knowledge.