A Computational Model for Spatial Reasoning with Mental Models

We propose a computational model for spatial reasoning by means of mental models. Our SRM model (Spatial Reasoning by Models) maps spatial working memory to a two-dimensional array and uses a spatial focus that places objects in the array, manipulates the position of objects, and inspects the array to find spatial relations that are not given in the premises. The SRM model results in a computational complexity measure that relies on the number of operations in the array and the number of relations that must be handled. The performance of the SRM model is compared to the performance of human subjects reported in the literature and in our own study