Authors Find AI Peer Reviews Useful but Flawed
A study of 56 authors at two computer science venues found that 83.9% considered AI-based peer review useful, and 80.4% reported it identified issues not mentioned by human reviewers. However, authors also noted inaccuracies and trust concerns. The research, published on arXiv, surveyed authors after review release using anonymous questionnaires, analyzing closed-ended items with descriptive statistics and open-ended responses with thematic analysis.
Key facts
- Study involved 56 analyzable responses from authors of 40 papers
- 83.9% of respondents found AI-based review useful
- 80.4% reported AI identified issues not mentioned by human reviewers
- Research conducted at two computer science venues
- Authors completed anonymous post-review questionnaire
- Closed-ended items summarized using descriptive statistics
- Open-ended responses analyzed using inductive thematic analysis
- Paper published on arXiv with identifier 2605.16623
Entities
Institutions
- arXiv