Title:Shaking the causal tree: On the faithfulness and minimality assumptions beyond pairwise interactions

Abstract:Built upon the concept of causal faithfulness, the so-called causal discovery algorithms have been proposing the breakdown of the contributions of the mutual information (MI) and conditional mutual information (CMI) to the sets of variables that reveal causal influences. These algorithms suffer from the lack of accounting emergent causes when connecting links resulting in a spuriously embellished view of the organization of complex systems. Here, we first show that causal emergence information is necessarily contained on CMIs. We also connect this result with the intrinsic violation of faithfulness. Finally, by using classical results from Ramsey's theory, we prove that CMIs with large conditioned datasets contain, necessarily, arbitrary correlations. These correlations appear only because of the size, and not the nature, of data. They can be found in "randomly" generated, large enough databases. We also connect this result with the false illusion of faithfulness. The net result proposes an update of the well-known causal discovery algorithms, which can, in principle, detect and isolate emergent causal influences in the network reconstruction problems undetected so far.