Choose interaction by investigation shape

These examples and illustrative results are independently authored teaching materials, not measured model results.

Use case

An intermittent queue stall may be lock contention or remote retries, unlike reviewing a known diff. Both teaching theories fit surface delay and need discriminating observations, not votes after isolated essays.

Mechanism

Determine known-artifact judgment versus causal discovery. Independent reports fit the former; the latter uses competing predictions, permitted evidence exchange and focused replay. Bound consensus to supported facts and retain indistinguishable explanations.

Bad example

Use three isolated theories, accept two votes and claim convergence without lock/call timing evidence.

Good example

Specify discriminating predictions for contention/retries. Exchange permitted wait/call/timing records and attempt falsification, then replay the same inputs or a minimal probe. Unresolved ambiguity remains two hypotheses, not majority truth. Known-diff independent review uses report collection separately.

Why the change matters

Investigation needs new evidence excluding explanations; review judges a fixed object. Collaboration should provide that value rather than replacing evidence with headcount.

Observable expectation

Repeated remote calls without lock waits support a retry direction but still need replay; delay alone discriminates neither. Inspect predictions/observations, not agreement logs.

Both approaches can fail; no real incident is investigated.

Limits

Discussion can anchor participants; retain initial hypotheses/counterevidence. Frozen host flags/versions or single-agent generalizations are not universal facts. Verify capabilities/cost and preserve tool authorization.