Causal Inference via Constraint Satisfaction
Amos Golan · 2017
Abstract In this chapter I introduce a number of ideas connected to causal inference that are inherently connected to info-metrics. In the context of this chapter, causal inference means the causality inferred from the available information. I begin by introducing and examining nonmonotonic and default logics, which were developed to deal with extremely high conditional probabilities. Other facets of info-metrics and causal inference are then discussed. I also show the direct effect of the complete set of input information on the inferred solution. I conclude the chapter with a detailed Markov example providing a more traditional approach to causal inference, developed within the info-metrics framework. The example builds on the notion of exogeneity and demonstrates that the info-metrics framework provides a simple way of incorporating additional exogenous information, thereby opening the way for empirical testing of causal inference. A short summary of the notion of “pure” causality is also provided.