A Bayesian Abduction Model for Extracting Most Probable Evidence to Support Sense Making

Abstract

In this paper, we discuss the development of a Bayesian Abduction Model of Sensemaking Support (BAMSS) as a tool for information fusion to support prospective sensemaking. Currently, BAMSS can identify the Most Probable Explanation from a Bayesian Belief Network (BBN) and extract the prevalent conditional probability values to help the sensemaking analysts to understand the cause-effect of the adversary information. Actual vignettes from databases of modern insurgencies and asymmetry warfare are used to validate the performance of BAMSS. BAMSS computes the posterior probability of the network edges and performs information fusion using a clustering algorithm. In the model, the friendly force commander uses the adversary information to prospectively make sense of the enemy's intent. Sensitivity analyses were used to confirm the robustness of BAMSS in generating the Most Probable Explanations from a BBN through abductive inference. The simulation results demonstrate the utility of BAMSS as a computational tool to support sense making.