Evaluating and refining computational models of spatial relations through cross-linguistic human-subjects testing

Abstract

Human-subjects testing based on a sound formal model is outlined as an effective way to evaluate and refine computational models of spatial relations. The cognitive response of a subject (person) to a real-world situation depends on the characteristic of the situation and the characteristics of the subject. Spatial entities can be characterized by topological relations, metric, scale (scope), kinds of phenomena (semantics), motion (time), and (for maps and displays) graphic presentation. Subjects' responses may vary according to their natural languages, their cultures, and individual characteristics such as gender, age, and handedness. Given this conceptual background, seven human-subjects testing protocols are introduced. Each test is described, and results of applications of these tests are summarized where possible. The tests are compared with their requirements, and the different aspects of human spatial cognition that they might test. Lastly, a program for applying these tests and refining the formal models based on test results is presented.