Computational imagery and diagrammatic reasoning: A case study in kinematics

We present KAP a computer program that solves planar higher-pair kinematic problems. KAP is a direct attempt to mirror human capacities to reason about kinematic problems using mental imagery. We implement the spatial and multi-scale nature of the visual buffer as a pyramid of occupancy arrays, and also present computational realisations of the attention window, cued attention shifts, the shape shift subsystem and the visual routines underlying the process of reasoning about object interactions. We contrast this approach to mainstream AI approaches to kinematic reasoning to demonstrate KAP's utility not only as a successful cognitive model, but also as an embodiment of a promising paradigm for a problematic class of spatial reasoning problems.