Establish whether model-written feedback improves learning in legal education

Establish whether feedback written by language models improves learning outcomes for law students compared with tutor feedback or other human-provided feedback.

Background

The paper considers formative feedback the use case with the strongest present justification for rubric-based judging. Rubrics and feedback have been associated with improved learning, and individual feedback has improved law students’ grades in prior work.

Nevertheless, the paper distinguishes evidence for feedback in general from evidence for feedback generated by LLMs in legal education. Students in German law have tended to prefer tutor feedback, and overly generous model grades may mislead students. The specific educational effect of model-written feedback therefore remains unresolved.

References

Whether model-written feedback improves learning in law is untested; students in German law still tended to prefer tutor feedback \citep{hackl2024kifeedback}, and generous model grades mislead, so feedback should state its uncertainty.

— SoK: Grading Open-Text Legal Answers, from Law School Exams to Trustworthy Rubric Judges  (2610.19992 - Nagl, 14 Oct 2026) in Section 5, “What the evidence supports,” subsection “Formative feedback”