Visual analytical approaches to evaluating uncertainty and bias in crowd sourced crisis information

Concerns about verification mean the humanitarian community are reluctant to use information collected during crisis events, even though such information could potentially enhance the response effort. Consequently, a program of research is presented that aims to evaluate the degree to which uncertainty and bias are found in public collections of incident reports gathered during crisis events. These datasets exemplify a class whose members have spatial and temporal attributes, are gathered from heterogeneous sources, and do not have readily available attribution information. An interactive software prototype, and existing software, are applied to a dataset related to the current armed conflict in Libya to identify `intrinsic' characteristics against which uncertainty and bias can be evaluated. Requirements on the prototype are identified, which in time will be expanded into full research objectives.