Confusion, social knowledge and the design of intelligent machines

Abstract

This paper explores the notion of confusion as it applies to the design of intelligent machines and suggests that it arises from the social nature of knowledge. It provides a framework for analysing the causes of confusion and shows how confusion normally reveals itself after the cause of the confusion has occurred. This places practical limitations on how intelligent machines can react to confusion when it is noticed. The paper also describes a prototype system that aims to minimize the effects of confusion by allowing the users to explore the design rationale of the system.