Echo risk

Agreement between models is not automatically evidence. Convergence is earned when reviewers reached the same answer independently, and social when they read each other and drifted into agreement. The distinction is measurable in a crude way — verdict flips toward the majority, overlapping citation sets, self-reported deference — and worth measuring, but the honest claim is narrow: it flags possible echo, it does not prove independence. High overlap can be an honest read of a small source.