Human Information Processing in Complex Networks

A curious aspect of information is its relativity: the amount of information contained in a message depends not just on its content, but also on the expectations of a receiver. Nowhere is this observation more evident, nor are the implications more important, than in the context of human cognition. Here, we develop an analytical framework for measuring information relative to human expectations, and we demonstrate its utility in two distinct ways. First, we verify that our framework predicts aspects of human behavior that cannot be accounted for by traditional information theoretical measures such as entropy. Second, we apply our framework to characterize the network structure of designed information sources, such as the network topology of natural languages and the structure of note transitions in music. Across a range of real-world networks, we discover that their inherent complexity is high while their divergence from human expectations is low, thereby allowing for the efficient communication of information. We find that this competition between high complexity and low divergence from expectations is driven by hierarchically modular organization, which, interestingly, has been observed in many evolved and designed networks.