Common sense reasoning about beliefs
Gezina Cornelia Stein, John A. Barnden · 1996
Reasoning about people's beliefs and their common-sense reasoning requires one to be aware of people's special reasoning characteristics. People are not logically perfect reasoners. People use di erent reasoning techniques, make mistakes, and sometimes do not make expected inferences. People's reasoning is resource-bounded, context-sensitive, and exible enough to deal with uncertain, incomplete or even partially faulty information. Further, reasoning about people's beliefs requires one to decide what beliefs to ascribe and what beliefs not to ascribe, since often one does not know all the other person's beliefs. Traditionally, computer systems that reason about (people's) beliefs use rules of inference, applied to general knowledge. Such approaches fail since they portray people as reasoners that are unrealistically powerful, while at the same time the approaches do not su ciently address people's capability to reason with uncertain, incomplete or incorrect information. Also, belief attribution as carried out by such systems is done in a very limited way. This dissertation presents an alternative approach based on case-based reasoning. Instead of reasoning with general knowledge, case-based reasoning reasons with instances of people's beliefs and their common-sense reasoning. Beliefs are now part of a bigger entity: a case. Since