Characterising Model-Based Reasoning

Abstract

This paper explores the fundamental nature of the technology known as Model-Based Reasoning. The central concept of representation is examined and the roles of prediction and explanation are emphasised. We argue that these three properties are essential elements of models that support Model-Based Reasoning and give examples to illustrate the distinctions. We show the importance of principled model design, guaranteed completeness of solutions and explicit, executable model structure. The modelling trade-off of Puccia and Levins is examined in order to characterise qualitative models and the benefits of MBR are presented. Finally, a series of challenges for the future are outlined.