Dartmouth Events

Abstract: 

AI agents will soon be as commonplace as smartphones, making sequences of interconnected decisions that impact human lives—from serving as decision support in healthcare to shaping educational paths for millions of students. A defining challenge for the future of AI is how to build agents that effectively operate in these human environments. 

In this talk, I discuss my research on rethinking and building AI agents from the perspective of human-centered reinforcement learning. I first motivate the central challenges by drawing examples from the field of mental health. Next, I demonstrate how human-centered design principles can guide the development of better-aligned AI agents. As a step toward enabling these agents to learn from diverse feedback modalities, I highlight our work on building benchmarks for learning from human feedback. Then, turning to the goal of AI transparency, I show how we can learn human-understandable decision-making policies. Together, this work illustrates how human-centered reinforcement learning is a valuable approach for developing AI agents that can learn from and for the people whose lives they impact.

Bio: Stephanie Milani is a final-year Ph.D. candidate in the Machine Learning Department at Carnegie Mellon University. Her research focuses on building reinforcement learning agents to address human-centered and use-case-inspired challenges. Stephanie is a 2024 Future Leader in Responsible Data Science & AI and Rising Star in Data Science. Her research has been published at top machine learning and human-computer interaction venues, including ICLR, NeurIPS, and CHI. It also received an outstanding paper award at the ICML MFM-EAI workshop and a best paper award at the NeurIPS GenAI4Health workshop. Outside of research, she received the CMU Machine Learning TA award, co-organized the MineRL international competition series at NeurIPS, and received the Newman Civic Fellowship for her service to computer science education.