Xiaoying Tu

Anonymous peer review

Anonymous peer review

Anonymous peer review

Co-submitted with Prof. Raja Sooriamurthi. The research objective is to build a GenAI-powered conversation agent (an "Intelligent Assessor") to simulate instructor- or TA-led interviews with students on their SQL query assignments, which is effective in evaluating students’ authentic learning but is hard to scale. This strategy aims to ensure students properly reason out their SQL codes, signaling genuine mastery of the material, rather than relying on unauthorized external help. Our high-level assessment strategy is the following: The student response for a take home assignment would be provided to the system. The system would then generate and pose questions to the student, in the spirit of how a human instructor may interview the student, to assess the authenticity and depth of learning. Following our initial proposal submission last fall, we have been working with Vincent Sha and his team in developing a proof-of-concept system. Over the winter break we wrote a paper summarizing our methodology of teaching SQL. This paper, some sample interview questions and answers, and a document listing common SQL development mistakes were provided to a Large Language Model (LLM). The prototype system under development on ChatNode is available for experimentation. This preliminary bot has shown promising results and potential to be further developed into the Intelligent Assessor.

Anonymous peer review