[2603.23322] Leveraging LLMs and Social Media to Understand User Perception of Smartphone-Based Earthquake Early Warnings

Statistics > Applications arXiv:2603.23322 (stat) [Submitted on 24 Mar 2026] Title:Leveraging LLMs and Social Media to Understand User Perception of Smartphone-Based Earthquake Early Warnings Authors:Hanjing Wang, S. Mostafa Mousavi, Patrick Robertson, Richard M. Allen, Alexie Barski, Robert Bosch, Nivetha Thiruverahan, Youngmin Cho, Tajinder Gadh, Steve Malkos, Boone Spooner, Greg Wimpey, Marc Stogaitis View a PDF of the paper titled Leveraging LLMs and Social Media to Understand User Perception of Smartphone-Based Earthquake Early Warnings, by Hanjing Wang and 12 other authors View PDF Abstract:Android's Earthquake Alert (AEA) system provided timely early warnings to millions during the Mw 6.2 Marmara Ereglisi, Türkiye earthquake on April 23, 2025. This event, the largest in the region in 25 years, served as a critical real-world test for smartphone-based Earthquake Early Warning (EEW) systems. The AEA system successfully delivered alerts to users with high precision, offering over a minute of warning before the strongest shaking reached urban areas. This study leveraged Large Language Models (LLMs) to analyze more than 500 public social media posts from the X platform, extracting 42 distinct attributes related to user experience and behavior. Statistical analyses revealed significant relationships, notably a strong correlation between user trust and alert timeliness. Our results indicate a distinction between engineering and the user-centric definition of system accurac...