2006 Graduate Workshop in Complexity and Computational Social Science

Student Projects

Each student began a research project during the two-week workshop. Below are brief descriptions of these various projects. These projects will form the basis for dissertation chapters and/or journal articles.

Nils Bertshinger, Complex Systems, Leipzig (bertschi@mis.mpg.de).

Nils considers the issue of information gained while sampling a search space. Using NK landscapes as the underlying search space, he derived a mutual information measure that captured the additional information needed to find the optimum. Using this procedure he analyzed the performance of learning algorithms across different spaces. He also applied the idea to understand better the impact of modularity in function optimization and learnability.

Aaron Bramson, Political Science, U. of Michigan (bramson@umich.edu).

Aaron is investigating the impact of neutral mutation networks on social science. In many social situations, different behaviors can lead to very similar rewards, for example, shaking hands or bowing as a way to greet one another. Such neutral mutations may allow diversity and divergence to flourish within a social system. Using a computational model, he is studying the development of cultural and moral systems.

Michael Findley, Political Science, U. of Illinois at U-C (mfindley@uiuc.edu).

Michael is interested in the resolution of civil wars, in particular, how do negotiation and implementation processes lead to stable resolutions that can avoid being "spoiled." The model has groups, with varying capabilities and resolve, attempting to create a lasting peace settlement in an environment confounded with random shocks. He finds that as agents alter their discount rates the system can exhibit very different behavior. Moreover, subtle changes in belief updating can result in large changes in system behavior.

Martin Ganco, Business Administration, U. of Illinois at U-C (mganco2@uiuc.edu).

Martin employs a NKC model to understand better industrial dynamics. There are very distinct patterns of firm performance in new industries, for example, preexisting (diversifying) firms behave quite differently than start ups in terms of entry timing and market share trends. He assumes that firms can be described as a set of coupled bit strings and uses a NKC model to tie firm choices to market outcomes. He finds that while firms with more internal epistasis tend to be at a disadvantage early on in the evolution of an industry, they often attain a distinct advantage in later stages.

Simone Giansante, Computational Finance, U. of Essex (sgians@essex.ac.uk).

Simone's work focuses on the development of trading networks in a simple model of exchange based on Wicksell's triangle. Agents must learn about what goods to offer and accept from one another during bilateral trades that take place within networks of trading partners. If the network is fully connected, a single good emerges as the medium of exchange. As connectivity decreases, the medium of exchange alternates across epochs. At extremely low levels of connectivity, more than one medium of exchange can exist simultaneously. When agents are allowed to reconfigure their links to other traders, he finds that the networks tend toward characteristic configurations where the majority of agents have only a few linkages to others while a few agents become quite connected.

Stephen Haptonstahl, Political Science, Washington U. (srhapton@wustl.edu).

Stephen is analyzing how communication can enhance group coordination. He uses an abstract model that allows information-processing agents to learn how to send and receive tokens that have no a priori meaning. Agents play a N-player Battle of the Sexes game where, while they prefer to coordinate with one another, each has a favored coordination point. Preliminary results indicate that this framework can begin to explain how communication can lead to coordination in groups, even when such coordination is difficult due to agent incentives and a challenging external environment.

Dominic Rohner, Economics, Cambridge (dr296@cam.ac.uk).

Dominic wants to understand the interaction of information, reputation, and ethnic conflict through the use of an analytic game theory model. In the model, agents interact with one another and attempt to trade. If an agents violates the terms of trade conflict ensues and reputations are altered via social networks linked to the underlying ethnic groups. He finds that this reputation-based view is able to link the onset of conflict to ethnic polarization, fractionalization, and segregation.

Mary Shenk, Anthropology, U. North Carolina (mshenk@unc.edu).

Mary is modeling marriage markets in South Asia. Of particular interest is the recent rapid increase in dowry payments that has been observed in India and Bangladesh. Her model incorporates parental choices for son's education and wealth and daughter's dowry based on the parent's own wealth and education level. Children, after being endowed by their parents, enter a marriage market and are matched based on their qualities. She finds that small increases in the rate of return to wealth can lead to rapid booms in dowry payments.

Hye-Jin Youn, Physics, Korea Advanced Institute of Science and Technology (visang@gmail.com).

Hye-Jin is exploring how cooperation can emerge in a growing population of agents playing the Prisoner's Dilemma. Incoming agents adopt the strategy of the fittest agent in the population and link to that agent and one other, randomly chosen, agent. In such a model, entering agents begin to adopt cooperative behavior at a very rapid pace (this result holds even if there is a small chance of strategic mutation). However, this system is extremely sensitive to migration (re-wiring), and defection can overwhelm the system.

Rimma Yusim, Economics, Brandeis U. (yusim@brandeis.edu).

Rimma is interested in understanding issues surrounding the emergence of firms, especially the driving forces behind the distribution of firm size observed across many economies. She assumes that firms face increasing returns to scale while workers must decide on how much effort to offer to the firm---workers find it easier to shirk in larger firms. Finally, she allowed hiring to be influenced by friendship networks. She finds that the average worker effort level in the population is stable, though the effort in the largest firms is quite variable. She also finds that aggregate patterns of industrial activity differ from models that do not consider networks in important ways.

John H. Miller , miller@santafe.edu.