Abstract

This dissertation focuses on systems that reason about people's beliefs and their reasoning, especially people using common-sense reasoning with common-sense beliefs. Such a system does not only have to reason about people reasoning with their own beliefs, but also about people reasoning about other people's beliefs . ;Such a system should be aware that people are not logically perfect reasoners. They are not logically omniscient, they apply a variety of reasoning techniques and they make mistakes. Their reasoning is resource-bounded, context-sensitive, and flexible enough to deal with uncertain, incomplete or even partially faulty information. People are able to ascribe beliefs to other people, which is often necessary when interacting with unknown people. In the past, these issues have been addressed only partially and insufficiently. ;Traditionally, systems that reason about beliefs use rules of inference, reasoning with general knowledge. This dissertation proposes an alternative approach: case-based reasoning. Case-based reasoning is a technique that uses specific previous experiences to solve new problems. It has been successfully used for different kinds of problems, but not for reasoning about beliefs. ;A system called CaseMent has been implemented to support this approach. The system combines case-based reasoning with a well-known technique called simulative reasoning. This results in the system portraying agents as case-based reasoners. Not only does the system have cases, agents can have cases too, so that cases can be nested. ;Although CaseMent certainly can be improved, its performance is encouraging. Its main advantages derive from its use of cases based on realistic instances of people's reasoning and beliefs, and its context-sensitivity, which is essential for many issues