Consensus, Divergence, and Outliers

Consensus items are suggested by many agents independently. Repetition indicates that several different analytical paths converged on the same conclusion.

Consensus does not prove correctness. Models can share the same blind spots or training patterns. It does make the idea important enough to examine closely.

Divergent items receive mixed support. Some agents recommend them, while others oppose them or prioritize something else.

This disagreement is valuable because it identifies the areas where the decision is sensitive to assumptions. The user should inspect the reasons rather than simply averaging the disagreement away.

Outliers appear in only one or a small number of responses. They may be hallucinations, irrelevant suggestions, or unusually strong ideas that the more common response patterns missed.

The correct response to an outlier is neither immediate rejection nor blind enthusiasm. It should be treated as a candidate for validation.

This classification allows a user to move from an undifferentiated list of suggestions to a structured decision landscape.