Evidence and limits
Finding: Psychological safety was associated with learning behavior, which mediated team performance in the studied setting.
Impact: Watch safe correction and escalation as human-workflow conditions. Reject interface affordances as proof that people feel safe to challenge outputs.
Boundary: One organizational setting does not establish causality across AI-assisted work. Next: how can an evaluation detect suppressed disagreement without collecting sensitive free text?
Flags: observational team evidence; organizational confounding
Record State: Catalog review recorded; not a performance endorsement
What was reviewed: Structure and relevance screened; not a full evaluation