Emotionally Enriched Feedback via Generative AI

Summary

A randomised feedback experiment separates improved perceptions from no detected improvement in work quality.

Initially 425 students were randomised; platform analyses used 381 consenting students, with a smaller survey sample.

Supportive feedback improved perceived usefulness and reduced reported anger, without a significant resource-quality difference. A tiny timing effect is not deeper learning.

Summarises the 2024 preprint; later journal version not fully compared, with no long-term retention or transfer measure.