What’s Happening? Elements of Commonsense Causation

Abstract

Knowledge of causation is an important part of commonsense reasoning. We use cause-and-effect analysis to understand everything from why we caught the flu to how to make a video recorder save our favorite TV show. Our facility can be characterized by a combination of two main capabilities: the ability to predict the outcome of a set of causative events; and the ability to explain given facts by postulating a set of determining causes.

Causal reasoning is most effective when combined with another type of commonsense reasoning, the assumption of normal conditions under which the reasoning is valid. While the existence of these assumptions has been recognized in the philosophical literature on causation, relatively little attention has been paid to the practical aspects of reasoning involving causation and normal conditions: for example, what happens when some causes affect the conditions on which other causes depend?

The field of Artificial Intelligence (AI) has evolved several techniques for formally representing and reasoning about conditions that we normally assume by default. In this paper we present an application of these methods in formulating a theory that integrates causal and default reasoning. The main structure of the theory is a default causal net representing the causal connections among propositions in the domain. The formal theory has been implemented using computational techniques available in AI; here we are mainly concerned with the structure of the theory, and its relation to other work in AI.