Background and Context: Auto-graders are praised by novice students learning to program, as they provide them with automatic feedback about their problem-solving process. However, some students often make random changes when they have errors in their code, without engaging in deliberate thinking about the cause of the error. Objective: To…

This study explores the potential of the large language model GPT-4 as an automated tool for qualitative data analysis by educational researchers, exploring which techniques are most successful for different types of constructs. Specifically, we assess three different prompt engineering strategies -- Zero-shot, Few-shot, and Fewshot with…