Debugging Environment for Complex Cognitive Architectures (DECCA)

Increasingly, scientists are proposing that a prerequisite for human-level intelligence be a more human-like approach to the underlying hardware platform. These future systems will enable a new generation of cognitive architectures. Such complex cognitive systems will require new tools to support their unique development requirements—especially for testing and debugging. The difficulties anticipated in testing and debugging complex cognitive systems stem from the intrinsic properties of distributed systems that are potentially parallel and adaptive. The solution is a debugging environment that supports a broad range of cognitive architectures, including those that are highly parallel and evolutionary. Specifically, the solution must: • Enable specification of expected behaviors to define the success of tests • Provide the ability to set complex query-based breakpoints • Enable fine control of multiple program flows within the system • Provide test coverage for non-deterministic systems • Support continued learning through redefinition of expected behaviors • Provide greater insight into hidden underlying processes through such techniques as sensitivity analysis and differential debugging Our proposed Debugging Environment for Complex Cognitive Architectures (DECCA) leverages our practical experience in debugging existing cognitive systems and extends our technologies to address the needs of future cognitive architectures.