[CS] Chacha Chen Candidacy Exam/Mar 3, 2025

[via cs]

This is an announcement of Chacha Chen's Candidacy Exam.
===============================================
Candidate: Chacha Chen

Date: Monday, March 03, 2025

Time: 11 am CST

Remote Location: https://urldefense.com/v3/__https://uchicago.zoom.us/j/9535186672?pwd=aXNDdEUzcCtHSGFjcTl3Q0xNcS9udz09__;!!BpyFHLRN4TMTrA!4H9FZpar6f9C7t0Cwr4tZE7xQAsM-p-qC4w93aAS48fdoK-fP5Dhh4e487-SF1T4xo9J0Fzq6je29O2JbD_B$

Location: JCL 298

Title: Towards Effective Human-AI Decision Making: Improving AI Adoption and Performance in Specialized Domains

Abstract: With the explosive progress of machine learning models, especially recent foundation models, these advanced systems are beginning to reshape our daily lives and workflows. This makes human-centered AI research critically important, which aims at building and improving AI models to better support human tasks, and ultimately improve human decision-making. My work focuses on improving human-AI collaboration through both behavior study and building better AI systems. We conduct human studies with domain experts in real-world tasks to understand the implications of advanced, human-level ML models—specifically how they are perceived and used by domain experts in medical diagnosis, with a focus on prostate cancer. Our findings indicate that domain experts are hesitant to adopt AI tools, and even when they do, they struggle to appropriately rely on AI. Additionally, we evaluate multimodal LLM applications in critical healthcare domains and have found that current state-of-the-art multimodal LLMs fail in specialized medical tasks and domains, highlighting a significant gap. To address this, we propose effective post-training mechanisms to enhance model performance in specialized domains. More broadly, my research contributes to the growing understanding of how AI is evolving from simple tools to sophisticated collaborators in knowledge work and specialized fields.

Advisors: Chenhao Tan

Committee: Chenhao Tan, Yuxin Chen, James Evans