Title: Empowering Humans in Human-AI Decision Makings
Due to recent advances in Artificial Intelligence (AI), AI models can surpass human performance in various tasks unprecedentedly and are rapidly integrated into systems that assist humans in making decisions. However, deploying such systems into the real-world requires understanding of the potential risks and challenges we might face. How do we interpret and explain AI models’ predictions while still be aware of their biases and weaknesses? I discuss my work that empowers humans to make better decisions with AI models through AI-backed interactive systems. I describe: (1) how humans make decisions with models, (2) how humans learn counterintuitive patterns from models, and (3) how humans make low-stakes decisions with models. I conclude by discussing my future research perspectives on improving human-AI collaborations through human-centered approaches.
Vivian Lai (https://vivlai.github.io) is a final year Ph.D. candidate in Computer Science at the University of Colorado Boulder advised by Chenhao Tan and James H. Martin. Previously, she received her B.S. in Information Systems from Singapore Management University. Her research lies at the intersection of Human-Computer Interaction (HCI) and Artificial Intelligence (AI) and aims to empower humans to make better decisions with AI models through AI-backed interactive systems. During her time as a Ph.D. student, Vivian has interned with Microsoft Research and Dataminr, and published papers at top-tier HCI, AI FATE, and NLP conferences and journals such CHI, FAccT, CSCW, and EMNLP.