1. Societal Consequences of Deployed AI:

Studying how AI systems shape human cognition, epistemic diversity, and value formation at scale. This includes understanding what meaningful help looks like in open-ended domains where success is not obvious, how values get encoded and evolve through continuous human-AI interaction, and how the balance between imitation and innovation shifts as AI becomes part of how people think and learn.

2. What Models Actually Know: Evaluation Beyond Surface Outputs:

Designing behavioral probes and controlled testbeds to evaluate whether AI systems internalize principles or pattern-match to surface outputs. Drawing on developmental psychology methodology to test causal, relational, and world-model reasoning in vision-language and embodied systems across single and multiturn settings.

3. Active Learning and Human-AI Co-Adaptation:

Analyzing how humans and AI agents acquire knowledge through active intervention versus passive observation, and how these learning modes differ in robustness and generalization. Examining how model behavior shapes human reasoning and decision-making in interactive settings, and how diverse human goals and values in turn influence system behavior over time.