Measuring Causal Invariance Formally


Invariance is one of several dimensions of causal relationships within the interventionist account. The more invariant a relationship between two variables, the more the relationship should be considered paradigmatically causal. In this paper, I propose two formal measures to estimate invariance, illustrated by a simple example. I then discuss the notion of invariance for causal relationships between non-nominal (i.e., ordinal and quantitative) variables, for which Information theory, and hence the formalism proposed here, is not well suited. Finally, I propose how invariance could be qualified for such variables.

Entropy, 23(6), 690.
Pierrick Bourrat
Pierrick Bourrat
Senior Lecturer & DECRA Fellow

My research interests include the various concepts deployed in evolutionary theory, causation, and the interplay between biological and cultural evolution.