Research
My work sits at the intersection of AI-supported STEM learning, science and engineering practices, learning analytics, and AI-enabled educational assessment.
Recent Work
This line of work studies how AI chatbots can support STEM and science learning, including chatbot-supported visual programming, learner self-efficacy, and evidence synthesis on whether AI chatbots improve science learning. The emphasis is on using conversational AI to support meaningful engagement with science and engineering practices.
This line of work develops AI-enabled approaches for assessing student-drawn scientific models while representing scoring confidence and uncertainty. It connects scientific modeling, automated scoring, and formative assessment so that model-based evidence can become more useful for feedback and classroom decision making.
This work develops a benchmark for evaluating how large language models respond to science misconceptions and student response distributions. It brings together natural language processing, science education, and AI evaluation to examine whether LLMs can recognize and respond to misconception patterns in student thinking.
This work uses the High School Longitudinal Study of 2009 (HSLS:09) to identify key predictors of U.S. students' long-term persistence in STEM pathways. The analysis combines nationally representative longitudinal data, supervised machine learning, and learning analytics to study persistence across the secondary-to-postsecondary transition.
Project Engagement
The GENIUS Center is an IES-funded research and development center focused on integrating generative AI with STEM+C teaching and learning. Its GenAgent platform supports middle-school science and engineering practices through AI agents designed for roles such as mentor, learning buddy, collaborative learning agent, and teacher assistant. Current work includes AI agent design, scientific argumentation, modeling, inquiry, student grouping, and learning analytics.
The PASTA project develops an automatically scored three-dimensional science assessment system that can organize student performance evidence and support instructional decision making. The project connects NGSS-aligned assessment tasks, AI-supported scoring, student feedback, and teacher-facing instructional guidance for middle-school science classrooms.
Previous Projects
This project studied personalized storytelling activities delivered by social robots to support children's English learning and attention. The work connected robot-assisted language learning, attention measurement, classroom implementation, quantitative analysis, and manuscript development.
This project developed a performance expectation framework for STEM education among Hong Kong primary and secondary students. The work drew on interview data and thematic analysis to clarify expectations for STEM learning and inform manuscript development.