Hyoungwook Jin

Large language models (LLMs) can empower educators to build pedagogical conversational agents (PCAs) customized for their students. As students have different prior knowledge and motivation levels, educators must evaluate the adaptivity of their PCAs to diverse students. Existing chatbot evaluation methods (e.g., direct chat and benchmarks) are either manually intensive for multiple iterations or limited to testing only single-turn interactions. We present TeachTune, where educators can create simulated students and review PCAs by observing automated chats between PCAs and simulated students.

This project builds a system that supports programming novices to abstract generalizable subgoals from code examples. We elicit abstract thinking by asking learners to write down subgoals common to interrelated code examples. We use code-generation AI models to create feedback on the generalizability and abstraction level of learners' subgoals by testing if the models can generate code examples from learners' descriptions.