STOC asks for AI-use disclosure but says it won't affect review — researchers ask what's the point
fortnow · x · 2026-09-28
The STOC conference requires authors to disclose how much AI they used, while its CFP states that AI tool usage "will not be weighed in the evaluation of the paper."
- Blogger Bill Gasarch lays out four disclosure scenarios: proofreading only, AI shortening proofs, human questions with AI answers, and AI-generated questions and answers — none of which affect acceptance
- Upside: authors have no incentive to lie about AI use. Downside: if the information changes nothing, why collect it?
- He rebuts Gemini's stated rationale (transparency, accountability, credit): the transparency argument is circular, authors are already accountable, and genuine credit attribution would mean allowing AI as authors
More from Safety
- Abundance Institute releases open-weight AI policy framework and primers for policymakers — neil_chilson · 2026-09-28
- May AI safety report looks prophetic after the Hugging Face incident — birchlse · 2026-09-28
- NVIDIA launches Open Agent Safety Platform for controlling what AI agents can do — nvidia · 2026-09-28
- Humans Are Reading Your Copilot Prompts — And The Reviewers Are Horrified — 404mediaco · 2026-09-28
- Cloudflare incident: skipped block zeroing leaked tenant data across 18 of 24 containers — arpit_bhayani · 2026-09-28
- Humans Are Reading Copilot Prompts — And They're Horrified — 404 Media · 2026-09-28