Cognitive and Probabilistic Models of Group Decision Making

We introduce an experiment designed to study trade-offs in collaborative decision making environments such as finding the best level of selectivity and abstraction in sharing information, and their impact on the time course and accuracy of group decisions. Two models of the experiment are presented: a cognitive model using the ACT-R cognitive architecture and a probabilistic argumentation model using Markov Random Fields (MARF). The cognitive model relies on memory mechanisms such as spreading activation, partial matching and blending to judge when to share information, which facts are relevant to a given question, and how to aggregate probabilistic evidence. MARF carries out real world reasoning after formal theory of human argumentation while at the same time being flexible to accommodate the deviations from the theory. MARF follows knowledge engineering paradigm aiming at reaching correct reasoning as much as possible. Representative results from the experiment are presented and compared to the results of the two models. Implications of the results and avenues for future work are discussed