If it is possible to provide a rationale for a specific output of an AI system, that system can best be described as: