Making Robots Conscious of Their Mental States

John McCarthy asks what self-knowledge a useful robot needs. The reviewed 1995 AAAI report proposes logical representations of mental states and actions.^[1]

Method and contribution

Formal examples explore introspection, reasoning about non-knowledge through relative consistency, and observing goals and constraints. A robot that recognizes insufficient knowledge may seek external information rather than infer an answer it cannot justify.^[1]

Alignment relevance and limits

Machine Introspection connects epistemic limits and goal reasoning to reliable delegation. That relevance is an interpretive connection, not a demonstrated modern safety method. McCarthy offers preliminary fragments; he does not claim an implemented system with all these capacities or a theory of human consciousness. Functional self-representation does not establish subjective experience.

The author maintained later revised versions. This entry distinguishes them from the original 1995 report; see Making Robots Conscious of Their Mental States.

Historical context

Making Robots Conscious of Their Mental States appears in AAAI Technical Report SS-95-05 (1995). Later web revisions are separate versions.^[1]