Dynamic Consensus Analysis of Social Media (DCASM) for Rapid Crisis and Disaster Response Information Discovery

The abundance of social media in crisis situations offers tremendous opportunity for filling the information vacuum that typically exists in HA/DR operations. However, current systems for extracting information are limited by human-in-the-loop processing of media, and do not provide automated methods for separating signal from noise, extracting information, and trust validation in order to create actionable intelligence. Furthermore, many computational approaches to extracting information do not leverage insights into human interaction and communication that have been uncovered by social scientists. We propose to apply social science algorithms, originally developed for ethnographic study, to the problem of extracting actionable intelligence from social media in disaster situations. A shared experience of reality, especially a profound reality such as a crisis scenario, should be reflected in observed message patterns. The proposed system will discover the latent"real"experience creating these patterns by adapting a number of rigorously validated analytic techniques. This will not only provide fully automated methods for processing and filtering social media data, but will also clarify collective understandings of reported reality for evaluating the trustworthiness of that information, for identifying meaningful aberrations in these reports, for effectively categorizing this information, and reliable intelligence for use by crisis response units.