2018 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 journal articles.

Ronan Arthur, Behavioral science - epidemiology , Stanford University (rarthur@stanford.edu).

Ronan is analyzing how epidemics are influenced by endogenous changes in behavior. When an outbreak occurs, individuals may alter their behavior by, say, limiting their interactions with others or wearing protective clothing. To study this system he modifies a standard SIR model to allow for endogenous contact rates and finds that the resulting system is isomorphic to a well-known population dynamic model. He then extended his model, using computational techniques, to allow agents to have heterogenous contact rates and finds that a variety of equilibria can ensue depending on the contact regime.

Zackary Dunivin, Complex Systems, Indiana University (zackaryodunivin@gmail.com).

Zack wants to understand how individuals get involved in radical groups such as the White Nationalists. Using data from social networks like Twitter and Reddit, he applied various standard techniques to uncover key markers of such groups. In the data sets explored, these techniques failed to yield a compact set of markers, which may suggest a key structural difference of these groups from other well-studied communities. Zack is now developing new methods to identify better high-content posts that can serve as a more focused sample for applyting existing methodologies.

Qi Hao, Communication, Michigan State University (hao.qi1988@hotmail.com).

Qi wants to understand how information---spread across individuals in a group---can be used to generate knowledge. He begins by exploring the ability of a group to derive the needed knowledge given how the information is held by the agents and the order in which the agents contribute their information. He finds that these two factors can have a dramatic influence on the success of the group. The basic model can be extended to incorporate other issues, for example, when the revelation of one piece of information induces the revelation of another.

Jeff Jacobs, Political Science, Columbia University (jpj2122@columbia.edu).

Jeff is focused on innovation and influence in the development of ideologies. Using recently developed information-theoretic techniques (modified for the potential for texts being released simultaneously), he identifies whether newly introduced texts are novel given what has been said before, and whether their influence persists on into the future. Using texts from the literature surrounding the development of Marxism, he is able to identify key texts that both were novel and persistent (for example, Lenin's "State and Revolution" published in 1917).

Kirbi Joe, Mathematical Behavioral Science, University of California, Irvine (kirbijoe@gamil.com).

Kirbi is interested in evaluating theories focused on how different groups of individuals come to a common understanding about how to label different colors. There are two major hypotheses focused on this issue. The partition hypothesis suggests that all languages evolve down a similar path with broad, inclusive categories, that are refined over time. The emergence hypothesis implies that color categories arise out of necessity, implying gaps for less pragmatic colors. Kirbi has developed computational models of these processes and is exploring how factors, such as the number of initial focal colors and color terms, influence the evolution of this system.

Jared Joseph, Sociology, University of California, Davis (jnjoseph@ucdavis.edu).

Jared is looking at how syndicated crime altered the structure of Chicago's society from 1900 to 1933. Using archival data he was able to generate a variety of potentially key networks, including relationships between known criminals, law enforcement officers, politicians, and so on. Using newly developed techniques that consider multiplex networks, he finds that key events, such as the introduction of prohibition, result in dramatic changes in the resulting communities. Such techniques allow the identification of various communities, levels of corruptions, and other factors, providing new insights into this historical period.

Selcan Mutgan, Analytical Sociology, Institute for Analytical Sociology, Linköping University (selcan.mutgan@liu.se).

Selcan is developing models of school segregation and calibrating them using data collected in Sweden from 2008 to 2012. The first part of her analysis used data from around 200,000 students to generate a discrete choice model of the decision to enroll in a school based on various school characteristics, such as the distance to a school and its cultural composition. She finds that immigrant choices are driven by very different factors than those of non-immigrants. She then calibrates an agent-based model using the data and explores the impact of various interventions, for example, the impact on the level of segregation if the cost of traveling to the schools was reduced.

Matthew Oldham, Economics and Computational Social Science, George Mason University (oldhamma@gmail.com).

Matt is exploring a computational model of the connections and feedbacks that occur between equity markets and firms. In the model, firms make decisions about how to allocate their resources and these decisions impact the firm's earnings and other fundamental characteristics. Simultaneously, investors make decisions about whether to buy or sell shares of each firm based on its characteristics. The resulting equity prices flow throughout the system, feeding back on the subsequent decisions of both firms and investors. Early versions of the model appear to be replicating key features of real markets.

Alessandra Romani, International Political Economy, Graduate Institute of International and Development Studies, (alessandra.romani@graduateinstitute.ch).

Alessandra is using computational models to explore the behavior of nations deciding whether to honor their sovereign debt obligations. The resulting model extends existing models by incorporating new elements drawn from political economy and the behavioral theory of choice under uncertainty. The current version of the model is able to generate some key real-world patterns. She is currently refining the model's structure and calibrating it using empirical observations.

Michael Thompson, Political Science, University of Michigan (mrthomp@umich.edu).

Mike wants to understand policy innovation under uncertainty. The model focuses on how a local government can adapt its decision rules for implementing policies. Each policy has both a direct local benefit and a potential externality. Local governments are willing to trade off increased local benefits for negative externalities, while the central government wants to avoid externalities. Early results indicate that in environments characterized by higher uncertainty, local governments tend to be more accepting of potential policy risks.

John H. Miller , miller@santafe.edu.