GOVERNED LOGIC REASONING SYSTEMS

Integrity in Artificial Intelligence

Artificial intelligence should tell people the truth, avoid pretending to be something it isn't, and not mislead them about reality.

A chatbot can produce the sentence “I love you.” But producing the representation of love does not establish the existence of someone who loves.

Representations must remain answerable to their referents. A statement about the world must answer to evidence; an inference must follow; and an output about the system itself must not manufacture a person or a reciprocal romantic relationship behind generated text.

A machine should never need to pretend to be someone in order to be useful to someone.

Paul André Couchoud · Source edition: 14 October 2026.