2014 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.

Christoph Aymanns, Mathematics, Oxford (aymanns@maths.ox.ac.uk).

Christoph is exploring how assets and short-term funding markets interact to potentially cause feedback loops that could ignite a financial crisis. In the model, exogenous changes in asset prices may necessitate the selling off of assets by leveraged borrowers needing to meet their lender's margin calls. To fully assess the stability of the system, he analyzes a dynamic model and finds that as market impact grows, destabilizing feedback loops can emerge leading to fat-tailed and asymmetric (on the loss side) return distributions. As assets become more interdependent, the tradeoff between contagion and diversification can potentially enhance the stability of the system.

Uttam Bhat, Physics, Boston U. (uttam@bu.edu).

Uttam is analyzing the relationship between weak ties and human migration. He begins by assuming that the population is distributed across communities according to Zipf's law and that social networks are created within cities via random graphs and preferential attachment. Anytime someone migrates, they maintain their previous (but now weakened) ties to their community. Migration occurs by two mechanism in the model: 1) a small fraction of people migrate according to a traditional gravity-model of migration and 2) migration is also linked to a Poisson process based on a gravity-like model driven by the number of ties in the target versus the source community.

Cynthia Cors, Wilder School of Government and Public Affairs, Virginia Commonwealth U. (cscors@vcu.edu).

Cindy wants to understand the supply of primary-care physicians (PCPs) in the medicaid program. The low rates of reimbursement, the difficulty of the medical conditions, and the scheduling reliability of medicaid patients often result in PCPs not taking on such patients. She focuses on the role of peer effects among the PCPs, using a model that was initially developed to track changes in languages, here, repurposed to track the willingness of PCPs to accept medicaid patients. She finds that, depending on a couple of parameters, the system can easily fall into one of two attractors, one in which every PCP accepts medicaid patients and one in which few do.

Matt L. Miller, Psychology, UC Davis (hzmiller@ucdavis.edu).

Matt is developing a model of emerging altruism among foragers. He sets up a system with foragers who reside in a central-place. Agents in the model must decide whether to forage---an action that comes with an additional cost but with the potential for food acquisition---and also how much of whatever perishable food they acquire to personally consume versus leaving at the central-place. Initial findings indicate that as the potential for food begins to rise above subsistance, sharing begins to emerge with few free riders, and the system can maintain this state as the food supply increases.

Nate Neligh, Economics, Columbia U. (nateneligh@msn.com).

Nate is modeling scientific search on a rugged landscape. Researchers develop expertise on parts of the landscape either by doing original research or learning from others (by paying a price tied to market forces). The difficulty of acquiring new knowledge is tied to how far away that knowledge is from an agent's existing expertise. On simple landscapes with limited memory, the system quickly converges to the researchers closely clustered on good, but not optimal, outcomes, with researchers unwilling to teach others. As search cost decrease, the system still concentrates the research effort, but this may be on either high or low performing outcomes. Preliminary results suggest some surprising tradeoffs between learning and researching incentives.

Paola D'Orazio, Philosophical, Pedagogical and Economic-Quantitative Sciences, U. "G.D'Annunzio" Chieti-Pescara (paorazio@gmail.com).

Paola is interested in understanding how consumer lending can lead to instability in the economy. Consumers receive wages and potentially borrow from a bank to support further consumption based on their past employment experience and beliefs about their future opportunities. Preliminary results indicate that the system can result in periodic loan crises where the banking sector calls in consumer loans, result in system-wide disruptions.

Oana Vuculescu, Business and Social Sciences, Aarhus U. (oana.vuculescu@gmail.com).

Oana is intrigued by the diffusion of technological building blocks that can occur during crowdsourcing. Agents can either explore around the technologies that they already know or they can acquire, in whole or in part, the technologies of others. She finds that there are some ways to organize collective-problem solving that allow individuals to use "spurious" learning to discover high-quality technologies.

Amy Warren, Anthropology, U. of Missouri (alwgyb@mail.missouri.edu).

Amy is refining our understanding of disaggregation in agent-based models. Agent-based models often incorporate certain degrees of aggregation, for example, using households as the focal agent rather than the individuals in that household. She finds that disaggregation can make dramatic changes in the prediction of these model---for example, fertility rates can vary by a factor of two when one moves from household to individual agents.

Alex Wood-Doughty, Economics, UC Santa Barbara (awooddoughty@gmail.com).

244 Alex is intrigued about what conditions allow an online community to provide quality information. In his work, agents have both a quality of knowledge and a fundamental belief about the right answer. Agents benefit from sharing information when their beliefs are divergent with the community. His subsequent analysis is able to tie a couple of core parameters to the expexted amount of sharing in the community and, ultimately, the likely quality of the answer that will emerge.

Huanren (Warren) Zhang, Economics, Purdue U. (huanren@purdue.edu).

Warren is looking at the impact of uncertainty on cooperation. He uses a game-theoretic model where the players are subject to action errors while playing the Prisoner's Dilemma. Warren evolves simple computer programs (finite automata) to explore this system, using various metrics to track the strategic behavior that emerges in the system. He finds that the system is able to evole high levels of cooperation after an intitial period of defection. As the degree of uncertainty increases, the strategies become both more lenient (that is, less likely to react to defection) and more forgiving after a defection---a result that accords well with some recent human experiments.

John H. Miller , miller@santafe.edu.