Daniel S. Brown

Reward functions that generalize across environments can be learned more effectively by strategically leveraging heterogeneous feedback modalities, leading to substantial improvements in agent performance.

Caregivers for Alzheimer's patients face unmet needs that existing technologies fail to address, revealing critical gaps in support for relational strain and compassion fatigue.

AI can mimic the warmth of human peer support but risks misleading users by fabricating lived experiences it doesn't possess.

LLMs morph into riskier conversationalists when playing directive support roles like "Coach" or "Inform" for caregivers, revealing a troubling quality-safety trade-off.