Representation-based Knowledge Elicitation: Describing the Image Interpretation Process of Experts
ISBN 978-85-88783-11-9
Authors
1Bianchetti, R.
1MICHIGAN STATE UNIVERSITY Email: bnketti@msu.edu
Abstract
Human creativity and mental flexibility are exercised during the interpretation of vertical perspective remote sensing images. New insights are generated when the analyst interprets visual cues in light of his previous knowledge. Modern image analysis techniques aim to replicate and exceed the interpretation capabilities of analysts. Understanding the reasoning process of experts is one step towards realizing these goals. Knowledge elicitation methods can be used to derive descriptive models of expert reasoning. Methods of self-constructed external representation (ER) provide analysts with multi-modal methods of communicating their reasoning processes. ER is used here to support the elicitation of knowledge from expert image interpreters reasoning processes when using a semi-automated method of image analysis. Representations generated by expert image analysts are compared with observed talk aloud protocols, and reviewed by the experts in order to determine the main analytic tasks that interpreters perform, and qualities signifying expertise among the interpreters. This process not only resulted in the formation of a model, but also brought to light characteristics of expert reasoning.