Giving Feedback on Interactive Student Programs with Meta-Exploration

Evan Zheran Liu, Moritz Pascal Stephan, Allen Nie, Christopher J Piech, Emma Brunskill, Chelsea Finn

Keywords: meta-reinforcement learning, education

TL;DR: We build a system that interacts with a student program to find bugs and provides feedback with near human-level accuracy by showing that finding bugs is a meta-exploration problem

Abstract: Creating interactive software, such as websites or games, is a particularly engaging way to learn computer science. However, teaching and giving feedback on such software is hard — standard approaches require instructors to hand grade student-implemented interactive programs. As a result, online platforms that serve millions, like Code.org, are unable to provide any feedback on assignments for implementing interactive programs, which critically hinders students’ ability to learn. Recent work proposes to train reinforcement learning agents to interact with a student’s program, aiming to explore states indicative of errors. However, this approach only provides binary feedback of whether a program is correct or not, while students require finer-grained feedback on the specific errors in their programs to understand their mistakes. In this work, we show that exploring to discover errors can be cast as a meta-exploration problem. This enables us to construct a principled objective for discovering errors and an algorithm for optimizing this objective, which provides fine-grained feedback. We evaluate our approach on a set of 700K real anonymized student programs from a Code.org interactive assignment. Our approach provides feedback with 94.3% accuracy, improving over existing approaches by over 17.7% and coming within 1.5% of human-level accuracy.

Community Implementations: [ 3 code implementations](https://www.catalyzex.com/paper/arxiv:2211.08802/code)