Crowdsourcing using Intelligent Supervision to address Information Requirements in Crisis Situations (CRISIS)

Marine Expeditionary Units (MEUs) are often first responders, addressing crises such as natural disasters and regional instability. Crowdsourcing provides a powerful potential tool to help MEU analysts collect and analyze information in new operational settings, but Marines need tools to manage crowdsourcing. MapReduce addresses a similar distributed parallel processing problem, without the diversity of human resources. Inspired by MapReduce, in our Phase I effort we designed and demonstrated a system for Crowdsourcing with Intelligent Supervision to address Intelligence Requirements in Crisis Situations (CRISIS). CRISIS provides a framework to employ diversely skilled crowds to address sensing and analysis problems for MEU analysts. CRISIS includes four key components founded in MapReduce functions: (1) human-centered problem partitioning to construct crowdsourcing problems; (2) computational models for understanding the relevant skills of workers; (3) a market-based optimization approach managing load management of potentially scarce skilled workers; and (4) solution reduction tools that aggregate results and synthesize understandable solution presentations for analysts. Using live and synthetic tests, we demonstrated an initial CRISIS capability that reduces error in crowd sensing problems, and provides critical information in crisis situations. Based on our Phase I success, we recommend a Phase II effort to develop and evaluate a full-scope CRISIS system.