Modelling a Fractionated System of Deductive Reasoning over Categorical Syllogisms

Abstract

The study of deductive reasoning has been a major research paradigm in psychology for decades. Recent additions to this literature have focused heavily on neuropsychological evidence. Such a practice is useful for identifying regions associated with particular functions, but fails to clearly define the specific interactions and timescale of these functions. Computational modelling provides a method for creating different cognitive architectures for simulating deductive processes, and ultimately determining which architectures are capable of modelling human reasoning. This thesis details a computational model for solving categorical syllogisms utilizing a fractionated system of brain regions. Lesions are applied to formal and heuristic systems to simulate accuracy and reaction time data for bi-lateral parietal and frontotemporal patients. The model successfully combines belief-bias and other known cognitive biases with a mental models formal approach to recreate the congruency by group effect present in the human data. Implications are drawn to major theories of reasoning.