Skip to main content
@shmVirus

publications

Where Automated Feedback Should Stop: Boundary Cases for LLM Support in Programming Assignments

Sabbir Hosen Mamun, Yuki Tanaka

Learning Analytics and Knowledge Workshop on AI-Supported Feedback, pp. 31-38, Dublin, Ireland · SoLAR

Conference2025

Abstract

Generative feedback systems can support learners, but they can also overstep by solving the task, masking uncertainty, or bypassing intended struggle. This workshop paper analyses 420 feedback interactions from an introductory programming course and proposes a boundary taxonomy for feedback granularity. The taxonomy distinguishes conceptual nudges, diagnostic explanations, implementation hints, and solution-equivalent disclosure.

Cite this work

@inproceedings{Mamun2025Where,
  title = {Where Automated Feedback Should Stop: Boundary Cases for LLM Support in Programming Assignments},
  author = {Sabbir Hosen Mamun and Yuki Tanaka},
  year = {2025},
  booktitle = {Learning Analytics and Knowledge Workshop on AI-Supported Feedback},
  pages = {31-38},
  address = {Dublin, Ireland},
  publisher = {SoLAR},
  abstract = {Generative feedback systems can support learners, but they can also overstep by solving the task, masking uncertainty, or bypassing intended struggle. This workshop paper analyses 420 feedback interactions from an introductory programming course and proposes a boundary taxonomy for feedback granularity. The taxonomy distinguishes conceptual nudges, diagnostic explanations, implementation hints, and solution-equivalent disclosure.},
}