Reasoning objects with dynamic knowledge bases

Abstract

Object-oriented programming has proven its appropriateness for simulating real wolds, in particular for imitating human societies and their ability to solve problems. Object-oriented software is easy to modify and to extend, a property of great importance for AI applications. Logic programming on the other hand stands out for its declarative specification language, built-in inference capabilities and clear theory. A well known feature of logic programming is the separation of knowledge representation and inference method.

We present a framework which amalgamates object-oriented and logic programming. It combines the object-oriented view with the logic formalism. Objects are considered as reasoning entities whose knowledge bases may change over time. They communicate via messages in order to ask for or to provide information. In response to new information, an object may have to update its knowledge. Operationally, reactions to messages are inference processes based on Prolog's inference by resolution mechanism. Great importance is devoted to simple and intelligible semantics of knowledge base alterations being the only way to change states. To this end, an object's knowledge base is divided into three parts: assumptions, reflections and reactions, each consisting of Horn clauses. Only assumptions are allowed to be altered. Knowledge can not be modified while an inference process is going on, resulting in easy-to-understand and easy-to-formalize semantics.