Bell Inequalities for Complex Networks.

摘要

This effort studied new methods to understand the effect of hidden variables affecting complex systems. Bell inequalities are a famous example of a hidden variable test in quantum physics that provides the strongest evidence for that theory. Initial work in this project extended the mathematical formulations of Bell inequalities to design new hidden variable tests that were able to account for confounding effects in complex systems including human social networks. These tests solved an open question about the identifiability of contagion in social network studies. Subsequent work moved beyond identification of hidden variables to develop a new information-theoretic framework capable of reconstructing hidden variables explaining the multivariate dependencies in complex systems. These methods have demonstrated value on diverse problems including human behavior, language, neuroscience, and gene expression.