The Innovative Educational Computing Lab is urgently hiring two Graduate Research Assistants (GRA) for a newly funded NSF project on the SimStudent project. Detailed project description can be found below. 

The GRA is only for PhD level graduate students.  The GRA scholarship will be available from Fall 2024 for three full years with potential continuation contingency upon future funding opportunities. 

The goal of the newly funded project is to develop a conversational pedagogical agent that functions as a synthetic peer for middle school students to learn to solve equations by teaching the synthetic peer.  Technologically, we aim to develop a computational model of question asking in natural language (aka an NLP dialogue management system) so that the synthetic peer asks constructive questions while the student is teaching.  We apply existing large language models and extend thier functionalities to derive expected dialogue between the synthetic peer and the student. From the learning science point of view, we aim to advance a theory of how students learn complex algebra knowledge by teaching. 

Ideal candidates should have prior research experience in machine learning, natural language processing, learning technologies, learning sciences, and/or other related disciplines. 

Interested candidate should send a current CV along with a research statement to Dr. Noboru Matsuda <>

Project Title: Learning by teaching with constructive tutee inquiry for robust learning in algebra

Summary: This project aims to advance the knowledge in how students learn robust knowledge in algebra that, by definition, allows students to not only derive answers for stereotypical problems but also draw analytical reasoning for unseen problems. Algebra is a gateway for broad STEM pathways. Yet, many students fail to achieve proficiency in algebra, which is arguably a primary cause of inability to pursue advanced STEM disciplines and further hesitancy in taking STEM pathways. The investigators hypothesize that one of the challenges in learning algebra is due to the complication of the web of algebraic knowledge students need to learn. It is argued that the web of knowledge involves conceptual and procedural knowledge and their relations, which the investigators call the connected knowledge. The investigators then propose to develop a transformative technology in the form of teachable agent to amplify the effect of learning by teaching that they call a smart teachable agent. The smart teachable agent asks students questions to justify their reasoning while solving equations. When student’s response needs to be elaborated, the smart teachable agent further provides a follow up question to solicit a response that reflects a connection between procedural operations and conceptual justifications. The smart teachable agent may ask follow-up questions two to three times. The proposed question-based dialogue between the student and the smart teachable agent is called the constructive tutee inquiry.

To implement the constructive tutee inquiry, the investigators will develop an innovative application of large language models (LLM) where multiple LLM invocations will be combined, including one for generating an ideal response to the agent’s question and another one for generating a follow up question based on the gap between the student’s response and the ideal response. The proposed dialogue system will be embedded into an existing online learning environment, called APLUS where students learn to solve linear equations by teaching a teachable agent, called SimStudent. As a learning science contribution, the investigators will study a theory of how students learn connected knowledge and how acquisition of connected knowledge facilitate robust learning in algebra. Classroom evaluation studies using APLUS and SimStudent with the proposed constructive tutee inquiry will be conducted with middle school students in their algebra classrooms.