【專題演講】LLMs as Measurement Tools: A Disaster Triage Case Study
Time: Thursday, November 27, 2025, 2:00 PM
Venue: Conference Room II, RCHSS, Academia Sinica
(128, Sec. 2, Academia Road, Nangang, Taipei)
Speaker: Aaron Lai
Aaron Lai is the Director of Analytics at Manifest MedEx and a Senior Fellow at Purdue University’s Krenicki Center. Formerly a leader at PwC, Castlight Health, and Blue Shield of California, he holds an MSc in Healthcare from Oxford and is completing a Doctor of Technology degree at Purdue, focusing on AI-enabled disaster triage. He is also joining the California State Guard to put this work into practice.
Abstract:
This presentation introduces a novel framework for computational social measurement using Large Language Models (LLMs). Using patient triage in disaster as a case study to illustrate the key challenges and mitigations, we demonstrate a multi-stage process that transforms unstructured text into standardized data for robust analysis. Our approach employs an LLM ensemble to not only perform the measurement task but also to rigorously validate the results, ensuring methodological reliability. This validation enables powerful policy simulations and underscores that different LLMs, like survey instruments, possess unique characteristics. We argue this framework offers a powerful, scalable tool for social science research, applicable to coding open-ended surveys and analyzing qualitative interview data.
Webex: https://reurl.cc/WO6ekO
Meeting ID: 2643 389 1344
Password: 1127
Registration: https://reurl.
Please register by 8:00 AM, November 24.
Registration is required for onsite participants only. Online attendees do not need to register.