Research

Leadership operating under intensive AI assistance faces a paradox: an abundance of analysis can erode judgment rather than protect it — sycophancy, over-reliance, homogenized reasoning. This paper introduces the construct of cognitive sovereignty — the collective capacity to maintain independent, calibrated, and falsifiable judgment — and specifies its four dimensions, erosion mechanisms, and conditions of observability. The protection mechanism is the cognitive contradictor, defined by four invariant principles, from which the paper argues a governance framework for boards. Theoretical companion to Calibrated Dissent. The N = 10 case is illustrative, not evidence of effectiveness.

A verifiable, anti-hindsight protocol for auditing the forecasts of an adversarial AI — the “cognitive contradictor”. Each forecast is sealed in a public, hash-anchored ledger before any outcome is observable, then resolved by a deterministic rule over canonical public sources and benchmarked against a matched base-rate null. Instantiated on a fixed, publicly pre-sealed pilot of N = 10 high-stakes European M&A forecasts; outcomes resolve over 2027–2030. OSF-preregistered.