Timely Targeting in Global Crises: A Depth Data Approach to Predicting Compliance From Public Social Media

Abstract

This study introduces a depth data approach for predicting individual compliance behaviours during the early stages of global crises, using the COVID-19 pandemic as a case study. Unlike large-scale behavioural data approaches that require extensive historical data, which is in most cases proprietary, or public voice data approaches that rely on explicitly posted crisis-related content, this approach leverages small-scale, publicly available digital footprints (such as Facebook public personal pages), to generate timely and actionable insights. Analysing a compact, consent-based dataset of 206 participants, we demonstrate that in-depth analysis of multi-dimensional social media traces can effectively predict compliance with key health measures during the pandemic. Our models outperform null and voice-based benchmarks, and the findings are operationally translatable into real-world targeting tools, such as Facebook’s advertising infrastructure based on users’ interests and exemplary user profiles. This approach offers a scalable and privacy-conscious solution for early-stage intervention, particularly when targeting non-voicing users is critical and large-scale behavioural datasets are unavailable. By prioritising timeliness, inclusivity and practical deployment, the depth data approach contributes a useful toolkit for early-stage crisis management.