Abstract:
We have found that implementing a metacognitive loop (MCL), which gives intelligent systems the ability to self-monitor their ongoing performance and make targeted changes to their various action-determining components, can play an important role in helping systems cope with the unexpected problems and events that are the inevitable result of real-world deployment. In this paper, we discuss our work with MCL-enhanced intelligent systems, and describe the ontologies that allow MCL to reason about, appropriately classify and respond to the perfomance anomalies it detects.