Research
What becomes stronger in the learner?
More time with AI is not necessarily more learning. Our research agenda asks what helps people understand, remember, explain, and act independently.
These are research directions and hypotheses, not claims of demonstrated outcomes.
AI, adolescence & social connection
How can personalized learning support strengthen confidence and agency without replacing human relationships? We are interested in the boundaries, roles, and conversational behaviors appropriate for young people.
Questions we are exploring
- When should AI answer, question, pause, or encourage a conversation with another person?
- Can feedback strengthen confidence without encouraging dependence on the system?
- How should we assess communication, learner agency, and social participation alongside academic performance?
Personal feedback. Classroom scale.
How can each learner receive useful feedback while their teacher gains capacity? We study AI as a supportive layer within existing classrooms, with educators able to review and challenge its observations.
Questions we are exploring
- What should adapt: feedback, questioning, pace, practice, or the form of explanation?
- How can teacher feedback change the support an AI system provides next?
- Can specialized agents coordinate practice, assessment, planning, and feedback more usefully than a single assistant?
- How do we measure the effect on educator workload as well as learner understanding?
The effort that makes learning happen
Convenience should not quietly remove the thinking through which understanding develops. We explore active recall, self-explanation, appropriate challenge, and focused engagement, including their use with young adults.
Questions we are exploring
- What changes when a learner attempts an explanation before receiving an answer?
- Can adaptive questions distinguish genuine understanding from familiarity?
- Which interaction patterns improve later retrieval and independent problem-solving?
- When should AI stop helping so that productive effort can continue?
Memory, privacy & classroom governance
Personalization benefits from context. Learners deserve boundaries around what is collected, inferred, remembered, and shared. We investigate privacy as part of learning design.
Questions we are exploring
- What is the minimum information needed for useful personalization?
- Can learning-state summaries replace unnecessary conversational history?
- Which information belongs with learners, families, teachers, or institutions?
- Where could local processing, limited retention, and governed human review reduce risk?