Peer Grading Study Reveals AI Trust Crisis in Education
Students fear AI in peer grading undermines authenticity and accuracy.
Deep Dive
Researchers analyzed 107 papers and 114 Reddit posts (plus 300 comments) to synthesize student views on peer grading, after using a fine-tuned Gemini 2.5 model to filter an initial set of 659 posts and 6,607 comments. Students see both benefits (learning, skill development) and drawbacks (unreliable grading, unfairness, stress). An emerging concern: AI use in peer grading may weaken trust. Eight mitigation strategies were identified, with instructor oversight playing the most central role. Accepted for ECTEL 2026.
Key Points
- Fine-tuned Gemini 2.5 model analyzed 659 Reddit posts and 6,607 comments as an initial relevance filter.
- 107 academic papers and 414 Reddit items (posts + comments) were synthesized, revealing both benefits (learning, skill development) and drawbacks (unreliability, stress).
- New finding: AI use in peer grading may erode student trust; eight mitigation strategies were mapped, with instructor oversight as most important.
Why It Matters
As AI automates grading, this study warns that trust must be managed alongside efficiency gains.