I flagged two research papers for fake authors and both were accepted as orals
AI-generated research papers and reviews are flooding major conferences, exposing how fragile and overburdened the current peer review and publication system has become. Commenters describe a surge in low-quality, citation-hallucinating “slop” incentivized by publish‑or‑perish culture, unpaid reviewing, and loose enforcement of AI-use policies, while editors increasingly experiment with AI-assisted review themselves. The thread raises wider concerns about academic incentives, reputational systems, and whether existing institutions can adapt without reverting to gatekeeping by elite networks or collapsing under the volume of automated output.
Automation of the Research Pipeline
- Commenters note that AI is now commonly used to write, review, and summarize papers, especially in ML/AI.
- Some see this as “automating humans out” of academic publishing; others argue it’s mainly automating the production of low-quality “slop.”
- There is concern that techniques in papers are now implemented by AI as well, further reducing human involvement.
Quality, “Slop,” and Limits of AI Review
- Many reviewers report a surge of low-effort, AI-written submissions with hallucinated references and shallow claims.
- Several argue current AI cannot reliably judge research quality; even many humans struggle with this.
- Some acknowledge AI has improved grammar and readability for non-native speakers, but say the content often remains vacuous.
Reputation, Social Scoring, and Accountability
- One proposed fix is stronger reputational systems or “social scoring” for authors and reviewers.
- Supporters see transparent consequences and reputations as necessary for functioning large-scale science.
- Critics warn such systems are easily captured by power, bias, and shifting norms of “good” and “bad” behavior, and may become dystopian.
Peer Review Incentives and Structural Problems
- Peer review is widely described as unpaid, overloaded, and often delegated to junior people.
- Some want reviewers to be paid; others note this could just incentivize AI-generated reviews.
- Suggestions include submission deposits that are forfeited for bad-faith slop, or more “shame” via de-anonymized submissions and reviews.
- Several argue the system was already “rotten” before AI; AI is just accelerating its breakdown.
Open Access, Citations, and Detection Tools
- Some blame closed-access publishing for making citation validation harder; others say existence of citations is already easy to check via DOIs and public indices.
- Tools and APIs for citation graphs and automated bibliography checking are mentioned; concerns arise that slop authors will use them to hide obvious errors.
- There is debate over bans and penalties for undisclosed AI usage and plagiarism; some call for multi‑year submission bans, others note enforcement is inconsistent.
Scale, Metrics, and the Future of Academia
- Massive submission counts at major AI conferences are seen as unsustainable; reviewers get many papers outside their expertise.
- “Publish or perish” is likened to judging programmers by lines of code; it drives volume over rigor.
- Some see the current academic model as failing and potentially replaceable; others warn against tearing it down without a better system ready.