A General Instance-Based Model of Sensemaking in a Functional Architecture

Abstract

This paper describes a general instance-based learning model of sensemaking in the context of geospatial intelligence tasks. Building upon a model previously described in Lebiere, Pirolli, Thomson et al. (2013), our model captures human performance across two tasks involving generating and updating likelihoods based on simulated geospatial intelligence. The model predicted human performance in such cognitive functions as generating and updating likelihoods based on incoming information, and in hypothesis/strategy selection and updating based on likelihoods taken in the context of experiences learned from prior exemplar. We then describe an initial attempt at a general instance-based model of decision-making capable of performing any task describable as a directed graph.