Linguistic Regularities in Pedagogical Feedback and Their Relationship to Error Diagnosis

Determine whether linguistic regularities observed in large language model–generated text appear in pedagogical feedback and characterize how these regularities align with a tutor’s ability to identify and correct student errors.

Background

Comparative studies have documented differences between human and LLM-generated text, including lexical choice and stylistic variability, but have not examined instructional feedback or its quality.

The authors explicitly highlight uncertainty about the presence of similar linguistic regularities in pedagogical feedback and how these patterns correspond to tutors’ effectiveness at identifying and correcting student mistakes.

References

Consequently, it remains unknown whether similar linguistic regularities appear in pedagogical feedback or how such regularities might align with a tutor’s ability to identify and correct student errors.

— Large Language Models Approach Expert Pedagogical Quality in Math Tutoring but Differ in Instructional and Linguistic Profiles  (2512.20780 - Abdulsalam et al., 23 Dec 2025) in Section 2, Related Work

Providing feedback that is both accurate and actionable therefore remains an open challenge, one our preliminary findings identify as a key area for improvement in DBWorkout.

— WIP: DBWorkout: A Gamified SQL Practice Platform to Support Formative Learning in Database Courses  (2610.01174 - Nizamani et al., 1 Oct 2026) in Introduction