Using ethnography to design a mass detection tool (MDT) for the early discovery of insurance fraud

We describe a Mass Detection Tool (MDT) for early detection of insurance fraud. Ethnography was used to specify needs and process, capture expertise, and design an interface for triggering fraud indicators while capturing unexpected anomalies detected by claims handlers. The MDT uses a dynamic Bayesian Belief Network of fraud indicators, whose weights are determined by how predictive each indicator is of specific types of fraud. The system uses automated knowledge updating to keep pace with dynamically changing fraud, adding new indicators that emerge from patterns of repeated anomalies.