Discovering a Decision Maker’s Mental Model with Instance-Based Cognitive Mining: A Theoretical Justification and Implementation

The purpose of this paper is to provide a theoretical justification for, and describe an implementation of, instance-based cognitive mining (ICM), a process that analyzes multiple decision instances using the inductive learning algorithms of artificial intelligence to generate a mathematical representation of the decision makerʹs mental models, explicitly relating how the decision maker implicitly selects and weighs key factors in making decisions within a specific problem domain. The foundation and justifications of ICM are based on three distinct literatures: 1) knowledge creation (mental models and knowledge externalization), 2) cognitive science (tacit knowledge and instance-based learning), and 3) artificial intelligence (data mining and inductive learning networks). We also propose an architecture that integrates several technologies to capture and express a decision makerʹs mental model, and we develop a prototype ICM software implementation. Finally, we describe a preliminary experiment that applies the ICM process to small teams of decision makers that tests (and supports) two hypotheses: H1: the ICM-derived mental model representation provides the mediating causal process through which the set of key factor values affects the actual decisions made by a team of decision makers; and H2a: the ICM-derived mental model representation is consistent with the teamʹs self-reported algebraic, directional, or tacit relationship^), and the teamʹs self-reported key factors; or H2b: any significant differences between the ICM-derived mental model representation and the teamʹs self-reported relationships or key factors are not consistent with the teamʹs actual decisions.