ABSTRACT
Algorithms are not the only abstract models with which computer scientists are concerned, and the modeling of physical events is not what they are always concerned with. The natural example on which to base an artificial rational agent is the human reasoner, but the full range of quality of human reasoning involving even simple common sense has proven enormously difficult to model. An illuminating attempt to unify these approaches to defeasible reasoning, including various versions of circumscription, default logic and autoepistemic logic was given by Shoham. It was not difficult for researchers to come up with an abstract model of the reasoning process they wished to automate, as they had been using the model themselves for years, namely predicate calculus. Insofar as human deductive reasoning is concerned with proving mathematical theorems, machine implementations are relatively straightforward since the required abstract model is both readily available and formal.
