Robb Willer Proposes Splitting Social Science Journals by Content Type as AI Reshapes Research

On October 22–24, Stanford sociologist Robb Willer initiated and led a discussion on how AI is reshaping the social sciences. His core argument: AI-assisted empirical research will be faster and more transparent and shouldn't be resisted, but space for speculative, exploratory, aesthetic writing must be preserved.

Confirmed

• Citing Brendan Nyhan, Willer argued that if AI optimizes social science papers to be extremely "technically correct" and faster to produce, the aesthetic, exploratory, speculative, ambiguity-tolerant writing and thinking many scholars cherish will be lost.
• His proposal: journals split by content type. Most social science journals would focus on falsifiable theories, data, and analysis, with any degree of AI support allowed in production and review to improve rigor and efficiency, plus 100% open data and analysis; separate channels would remain for speculative, exploratory, subjective writing, distinguished through clearer labeling.
• He argued that preserving speculative and exploratory space benefits rigorous empirical work itself—one can explore a vast space of ideas with minimal constraints—and noted the real problem that unfalsifiable content and falsifiable empirical work are hard to coexist in the same venues.
• In closing remarks, Willer said he expects discoveries and insights to arrive at unprecedented speed, so we need to preserve shared spaces for human digestion—sharing work with each other and allowing imperfect, incomplete, not-yet-testable ideas to circulate freely.
• Economist Barbara Biasi (relayed by Jessica Hullman) offered a game-theoretic observation: when everyone uses AI (which she writes as Pgram) to read every paper, one possible equilibrium is authors first generating content with AI (slop), then spending time editing text to lower their AI score (Pscore)—often more time-consuming than not using AI at all. Hullman questioned whether that's a good equilibrium.

Why it matters

• The discussion addresses a structural question of academic publishing in the AI era: not simply banning or permitting AI, but using journal specialization to preserve both the speed and transparency of empirical work and the survival of speculative writing.
• The "AI slop game" posed by Biasi and Hullman reveals how detection scores interact with writing behavior: when lowering an AI score becomes the goal, incentives can warp—a warning for any publishing policy that relies on AI detection.

2026-10-22 ~ 2026-10-24 · 6 related posts

• [source] When everyone lowers their AI slop score, is that a good equilibrium? — JessicaHullman · 2026-10-22
• Sociologist warns AI-optimized papers could erase exploratory thinking in social science — RobbWiller · 2026-10-24
• Willer: AI-assisted positivist social science will get faster and clearer — RobbWiller · 2026-10-24
• Journals could split: positivist AI-assisted papers vs speculative writing — RobbWiller · 2026-10-24
• [source] Stanford sociologist: AI-optimized papers shouldn't crowd out speculative writing — RobbWiller · 2026-10-24
• [source] Willer: as AI speeds up findings, humans need room to process them — RobbWiller · 2026-10-24