Thursday, February 10, 2005
4:15 pm
B17 Upson Hall

Michael Macy
Cornell University

Group Dynamics in Silico: Complex Networks and the Human Flock

What do flocks of birds, traffic jams, fads, forest fires, riots, internet search engines, and residential segregation have in common? The answer is self-organization. There is no leader bird who choreographs the dance-like movement of a flock of geese. There is no supervisor in charge of a riot. There is no librarian in a back room at Google headquarters who is busily classifying all the internet websites in a digital version of the Dewey decimal system. There is no conspiracy of banks and realtors who are assigning people to ethnically homogenous neighborhoods.

Traditionally, sociologists have tried to understand social life as a structured system of institutions and norms that shape individual behavior from the top down. In contrast, a new breed of social modelers suspect that much of social life emerges from the bottom up, more like improvisational jazz than a symphony orchestra. People do not simply play parts written by elites and directed by managers. We make up our parts on the fly. But if everyone is flying by the seat of their pants, how is social order possible? New and compelling answers to this question are being uncovered by social theorists using an innovative modeling tool developed in computer science and applied with impressive success in disciplines ranging from biology to physics -- agent based computational modeling. ABC models are useful tools for exploring the complexity of interaction among interdependent decision-makers, identifying the attainability of equilibria, and uncovering the effects of network structure on population dynamics. Working with graduate students Damon Centola and Robb Willer, I demonstrate the advantage of this approach in an application to a social enigma based on Hans Christian Andersen’s classic fable, "The Emperor's New Clothes." In this model, agents must decide whether to comply with and enforce a norm that is supported by a few fanatics and opposed by the vast majority. We find that cascades of self-reinforcing support for a highly unpopular norm cannot occur in a fully connected social network. However, if we limit agents’ horizons to their immediate neighbors, highly unpopular norms can emerge locally and then spread throughout the population. One might expect these cascades to be more likely as we increase the number of “true believers” and create bridge ties between otherwise distant actors. Surprisingly, we observed quite the opposite effects. Explanations of these anomalies generate new and important insights into the spread of unpopular norms.