On Effective Causal Learning

Abstract

We have developed a framework for identifying causal relationships between events which are effective in the sense that they can be put to practical use, without regard to what the “true” causes really are. A rapid causal learning process is devised for temporally correlated events that can be observed proximally which is sufficient for the learning of many causalities involving basic physical and social phenomena. The system relies on a diachronic aspect of causes which is a characterization of consistent temporally correlated events and a synchronic aspect of causes which is a characterization of the contextual factors that enable the diachronic causal relations. The causal learning method is applied to some problem solving situations that allows some basic knowledge to be learned rapidly and that results in drastic reductions of the search space and amount of computation involved. This method is necessary to jump start the chain of causal learning processes that allow more complex and intricate causal relationships to be learned based on earlier knowledge.