Diagnostic reasoning from a casual model

Abstract

People seem to reason about some kinds of problems by conceptualising their knowledge as models. This thesis describes a system called MOBY which has a model-based representation of wheat plants and their diseases, and which reasons about the models to diagnose specimens. There are two key ideas in the thesis. The first is that MOBY's models explicitly represent biological processes that are relevant to plant diagnosis. The second is that because the models only specify disease mechanisms and not a way to solve problems, MOBY's inference engine must be more sophisticated than that of an equivalent rule-based system. To demonstrate the viability of the concepts described in this thesis, MOBY has been partially implemented.