Predicting spatial belief reasoning: comparing cognitive and AI models

Spatial relational descriptions in everyday life sometimes need to be revised in the light of new information. While there are cognitive models for reasoning about spatial descriptions there are currently no models for belief revision for the spatial domain. This paper approaches this need by (i) revisiting existing models such as verbal model (Krumnack et al., 2010) and PRISM (Ragni and Knauff,2013) and adapt them to deal with belief revision tasks, (ii) evaluate these models by testing the predictive accuracy for the individual reasoner on a previously conducted experiment by Bucher et al.(2013), (iii) provide baseline models and machine learning models, provide user-based collaborative filtering and content-based filtering methods, and provide an analysis on the individual level. Implications for predicting the individual and identifying strategies and shared similar reasoning patterns are discussed