Using Model Theory to Specify AI Programs

This paper proposes a method for adapting the traditional devices of model theory to the task of specifying the input/output behavior of artificial intelligence reasoning programs when viewed as inference engines. The method is illustrated by specifying two programs, one a toy example and the other a program for retrieving information from a declarative knowledge base. Close examination shows that many intuitions about the properties of a retriever can be stated rigorously in terms of inference and that the model-theoretic specification can then be used to prove that the retriever has these properties.