Learning Synthesis Schemes in Intelligent Systems

We present a settting in which one can discuss problems of design, synthesis, analysis and control ofcomplex systems by adaptive teams of intelligent distributed agents. We point to learning problems of this approach related to the necesity of extracting from the empirical data of constructs allowing the agents to negotiate their co operative actions. We put our analysis into the framework of multistrategy learning (Michalski 1994) which combines empirical induction, abduction and reasoning by analogy in a hierarchical setting (Michalski 1994).