A Computational Model of Risk-Context-Dependent Inductive Reasoning Based on a Support Vector Machine

A computational model of cognitive inductive reasoning that accounts for risk context effects is proposed. The model is based on a Support Vector Machine (SVM) that utilizes the kernel method. Kernel functions within the model are assumed to represent the functions of similarity computations based on distances between premise entities and conclusion entities in inductive reasoning arguments. Multipliers related to the kernel functions have the role of adjusting similarities and can explain rating shifts between two different risk contexts. Model fitting data supports the SVM-based model with kernel functions as a model of inductive reasoning in risk contexts. Finally, the paper discusses how the multipliers for kernel functions provide a satisfactory cognitive theoretical account of similarity adjustment.