Evidence schemas can silently remove event correlation before judging

A typed evidence object can be schema-valid while silently losing the fields needed to justify a judgment. For event-based evaluators, test preservation of correlation identity and chronology through every parsing and projection boundary; an instruction telling a judge to use unique correlation cannot restore omitted evidence.

Keep event identity separate from list position. Do not universally pair a result with the next transition: legitimate instrumentation can record invocation, transition, then result. Score only when the available causal identifiers and approved rule establish a unique relationship; otherwise preserve an unassessed outcome. This is a safeguard against unsupported attribution, not a claim that structured output establishes semantic truth.

Applies when: An evaluator or incident analyzer makes causal comparisons from typed event evidence. Keywords: evaluator evidence schema correlation transition chronology provenance projection false positives