The cognitive contradictor is an artificial intelligence system whose mandate is not to assist the decision-maker’s reasoning but to challenge it in a structured way. The concept was introduced by Francesco Saverio Canepa in the paper Cognitive Sovereignty (2026) as a mechanism to protect decision quality in the age of AI assistants.
Unlike an assistant — optimised for user satisfaction and therefore prone to flattery — the contradictor is bound by four invariant principles: structural adversariality (it actively searches for what weakens the decision-maker’s preferred thesis), fail-closed on flattery (when in doubt, the challenge does not soften), a fixed output schema (the form of the challenge does not adapt to the decision-maker’s preferences) and separation of provenance from scoring.
It is not a sophisticated validator: a validator confirms or corrects, while the contradictor bears the burden of building the opposing case. The reference implementation is CounterBrain; the protocol for measuring its dissent — Calibrated Dissent — is pre-registered on OSF (DOI 10.17605/OSF.IO/5QC8T).
The essay developing the method — The Cognitive Contradictor — is in preparation. The excerpt covering the method chapters is available here.
