"A Proposed Definition of Causal Information"

abstract: Claude Shannon taught us how to measure the amount of information that one random variable reveals about another; however, he did not differentiate between random variables that are causal and those that are merely statistical in nature. How can we quantify the extent to which such information is causal, and what exactly is causal information in the first place? In this talk, by bringing together Shannon's information theory with Judea Pearl's do-calculus, I will propose a definition of causal information, and if there is time, I will discuss potential implications for issues of social justice, and discrimination. In the first half of the talk, I will review basic concepts in information theory and do-calculus so that hopefully everyone will be up to speed.