Research

Current Research

My research develops human-centered AI that senses, models, and supports learning. I combine computer vision, speech and language processing, and generative AI to understand how students engage, regulate their learning, and collaborate. My work spans four interconnected themes:

1. Multimodal Learning Analytics and Affective Computing. Detecting and modeling learner emotions, cognition, and engagement in classrooms and open-ended learning environments using video, audio, eye gaze, speech, and interaction logs. Part of the NSF AI Institute for Engaged Learning (Engage AI), with papers at AIED, ICMI, CSCL, LAK, EDM, and ICCE.

   1. on human-in-the-loop AI for analyzing embodied learning in classrooms (Learning and Instruction, 2026)

   2. on relating students' cognitive processes and learner-centered emotions (ICMI 2024)

   3. on emotions, cognitive actions, and performance in collaborative learning (CSCL 2025)

   4. on LLMs for detecting socially shared regulation of collaborative learning (LAK 2026)

2. Responsible AI in Education: Privacy, Fairness, and Bias. Privacy-preserving de-identification that removes student identity while retaining pedagogically relevant emotion signals, and systematic audits of bias and robustness in emotion recognition models used in classrooms.

   1. on identifying and mitigating algorithmic bias in student emotion analysis (AIED 2024, Best Paper Nomination)

   2. on valence-arousal guided de-identification that retains emotion signals (ACM SAC 2026)

   3. on challenges of applying computer vision for emotion detection in education (AIED 2025)

3. AI for Mixed-Reality Training and Simulation. Video-based performance assessment, human-object interaction analysis, and teamwork evaluation in mixed-reality medical and military training environments, funded by the U.S. Department of Defense.

   1. on video-based performance evaluation for ECR drills (I/ITSEC 2025, Best Paper Nomination)

   2. on trainee action recognition in CCATT mixed-reality training (I/ITSEC 2025, Best Paper Nomination)

   3. on human-object interaction and gaze analysis for performance assessment (MHSRS 2026)

4. Datasets, Tools, and Open Science. Public multimodal datasets and open-source tools that let other researchers study learning at scale.

   1. a multisensor dataset of students solving mental rotation tasks (Scientific Data, 2025)

   2. an affective database of students' faces, gestures, and postures (FGCS, 2020)

   3. DLOT, an open-source observation and annotation tool (ICCE 2023, Best Technical Design Nomination)

   4. SyncFlow, a real-time multimodal analytics platform (US-RSE 2025)

I collaborate with learning scientists, K-12 teachers and schools, and government and industry partners. If our interests overlap, I would be happy to talk - please reach out via the Contact page.