Less IS More! Using Logic to Get Less Inference

Abstract

I will begin with background material on the connection between logic, on the one hand, and algorithms that draw inferences, on the other. This allows us to design inferential algorithms using the framework of applied logic. I will then argue that less inference is often better than more inference when working in numerous application areas, even though the intuitive meaning of "intelligence" typically requires more rather than less inference. Next, I will survey several techniques for capturing limited inference. Finally, I'll show how we do it in the specific area of planning to achieve given goals.