A Guide to Using Social Media as a Geospatial Lens for Studying Public Opinion and Behavior
New research framework uses LLMs to turn social posts into real-time maps of human behavior.
Researcher Lingyao Li has published a comprehensive guide on arXiv titled 'A Guide to Using Social Media as a Geospatial Lens for Studying Public Opinion and Behavior.' The paper presents a practical framework for treating user-generated content from platforms like X (formerly Twitter) and review sites as a form of passive, human-centered sensing. This method aims to complement traditional surveys and physical sensor networks by providing timely, spatially-anchored insights into public sentiment and behavior.
The guide outlines a four-step methodological workflow: platform-aware data collection, information extraction using advanced large language models (LLMs), geospatial anchoring of the data, and statistical modeling. It highlights how modern LLMs significantly strengthen the ability to parse meaning and structured information from the inherently noisy and unstructured text of social media posts. The framework's utility is demonstrated through four concrete case studies, including tracking COVID-19 vaccine acceptance and conducting rapid, approximate damage assessments after earthquakes.
This approach transforms scattered online chatter into a mappable dataset, enabling researchers and analysts to measure public attitudes with geographic specificity and near real-time speed. It argues for social data as a critical tool for understanding place-based experiences, from airport service quality to urban accessibility, at a scale and granularity previously difficult to achieve.
- Framework treats social media as a passive geospatial sensor, complementing traditional surveys and physical data.
- Leverages large language models (LLMs) to extract structured, analyzable information from unstructured social posts.
- Validated with four case studies, including real-time tracking of vaccine sentiment and post-disaster damage assessment.
Why It Matters
Enables organizations and governments to gain real-time, location-specific public sentiment insights for crisis response and policy planning.