Economics journals are piloting Refine, an AI that scans papers and appendices for mistakes; its creators say it found problems in roughly a third of already‑refereed papers. If adopted widely, such tools could change referee workloads, raise the bar for reproducibility, and shift editorial responsibility toward automated checks.
— Widespread use of AI in peer review would reshape scientific credibility, publication incentives, and how errors or 'sloppiness' are discovered and punished across disciplines.
Scott Alexander
2026.10.29
80% relevant
Scott reports using Claude to detect substantive errors in a marriage‑statistics piece (and recommends asking an AI about surprising claims), which illustrates the emerging practice of using large language models as instant auditors/fact‑checkers for journalistic link curation and commentary.
Dan Williams
2026.10.27
65% relevant
The piece highlights experiments and the potential for chatbots to improve accuracy and information quality, connecting to earlier themes about AI taking on evaluative, curatorial roles in scholarly and public evaluation processes (actor: chatbots used in thousands‑participant experiments mentioned).
Arnold Kling
2026.05.04
60% relevant
Kling discusses assigning an AI project to score opinion writing and Scott Alexander questions whether users will adopt platforms that privilege logical accuracy — this concretely connects to debates about using AI to automate evaluation and moderation (peer‑review/adjudication) of public discourse.
Tyler Cowen
2026.03.19
100% relevant
Refine (AI reviewing tool) being trialed by several top‑five economics journals and Ben Golub’s statement that Refine found issues in at least one‑third of cases.