Collaborative Filtering of spatial-temporal information for crisis informatics

In a disaster, accurate information is a resource that is often in short supply. The combination of rapidly unfolding events and the frequent loss of communication infrastructure including mobile phones, landlines, Internet and television broadcast make it difficult to gain situational awareness. And while there is obvious value in aggregating information for centralized emergency authorities, disasters like Hurricane Katrina have shown that authoritative organizations are not always prepared to mobilize quickly enough. There is significant potential for systems that support self-organization and information sharing among the victims of a disaster. This paper discusses an application designed to share disaster information over an ad-hoc network built from common mobile devices (e.g. smart phones and PDAs). A Collaborative Filtering (CF) algorithm aggregates the observations of numerous distributed users and disseminates maps of disaster related events back to them. This paper describes a framework for analysis of disaster events, including both the availability of necessary resources (e.g. water and fuel) as well as disruptive events (e.g. downed power lines, leaks, etc). In addition, a spatial-temporal clustering algorithm combines the observations of multiple users into coherent reports. The goals of the framework are to quickly spread and accurately update vital information while resisting rumors and errors. The results demonstrate that sharing information in this way can increase the group´s situational awareness.