A new generation of open and commercial AI tools is moving from assistant roles to evaluators of scholarship—flagging assumptions, mapping literatures (240K‑paper graphs), and offering model‑level critiques that could substitute for or reshape peer review. These systems lower the cost of meta‑research, but also concentrate power around tool builders and the signals their analyses produce.
— If AI takes on an evaluative gatekeeping role, it will reshape incentives, hiring, publication, and what counts as credible evidence in science and policy.
Arnold Kling
2026.04.16
75% relevant
Arnold Kling’s experiment — uploading a Claude 'skill' that tutors students on readings and surfaces the instructor’s perspective — is a direct instance of AI tools changing how students prepare, what counts as classroom‑ready work, and who controls assessment and onboarding of ideas (actor: Arnold Kling; platform: Claude; artifact: .skill reference file and live test link).
Tyler Cowen
2026.03.23
88% relevant
Tyler Cowen argues that AI services (e.g., 'Refine') will judge, rewrite, and continually improve past and new economics papers, shifting evaluation from journals to AI-driven assessment of datasets and code; this directly maps onto the existing idea that AI tools will change how scholarship is evaluated and rewarded (tenure, prizes, gatekeeping).
Tyler Cowen
2026.03.19
100% relevant
The article links to 'Show Me The Model' (an AI that flags hidden assumptions) and Frontier Graph (an open tool built from 240K papers), showing this is already happening.