News Category: ‘Past Friday Seminars’

Friday, April 29 – 3:00 pm

Bringing the Con Back to Text: Using Contexts for Multiagent Modeling of Complex Environments

Seyed Mohammad Mussavi Rizi
PhD Candidate, CSS
George Mason University

Abstract: In this talk I will present a design pattern that helps to deal with complex agent behaviors in socio-natural multiagent models. This design pattern uses the notion of context behaviors to divide agents’ complicated decision cycles into portions that are intuitive and lend themselves to more robust coding. I will demonstrate context behaviors by demonstrating how it was used in a complex model of Afghan agriculture.

Tuesday, May 3 – 2:00 p.m.

How to Use Decision Theory to Choose Among Mechanisms

David Wolpert
NASA Ames

We extend a recently introduced approach to the positive problem of game theory, Predictive Game Theory (PGT Wolpert (2008)). In PGT, modeling a game results in a probability distribution over possible behavior profiles. This contrasts with the conventional approach where modeling a game results in an equilibrium set of possible behavior profiles. We analyze three PGT models. Two of these are based on the well-known quantal response and epsilon equilibrium concepts, while the third is entirely new to the economics literature. We use a Cournot game to demonstrate how to use our extension of PGT, concentrating on model combination, modeler uncertainty, and mechanism design. In particular, we emphasize how PGT allows a modeler to perform prediction and mechanism design in a manner that is fully consistent with decision theory. We do this even in situations where conventional approaches yield multiple equilibria, an ability that is necessary for a fully decision theoretic mechanism design. Where possible, PGT results are compared against equilibrium set analogs.

Friday, May 6 – 3:00 pm

Agent-based Simulation of Tax Reporting Compliance

Kim Bloomquist
PhD Candidate, CSS
George Mason University

Abstract: I describe the development of the Individual Tax Compliance Model (ITCM), a hybrid multi-agent simulation (MAS) test bed for evaluating alternatives to improve income tax reporting compliance. ITCM simulates the tax reporting behavior of individual taxpayers, including their formal economic relationships (e.g., employers, commercial tax preparers) and informal reference groups, for a mid-sized community in tax year 2001. In all, the full model has nearly 85,000 taxpayers, 3,300 employers, 2,100 tax preparers, and 21 geographic zones.

To comply with IRS rules on nondisclosure of taxpayer data, while also permitting external model validation and verification, I create a dataset of artificial taxpayers by replacing actual tax returns with cases from the Statistics of Income (SOI) Public Use File. To this dataset I impute income and offset misreporting using random audit data from the National Research Program. I calibrate the model to tax return data and random audit results.

The model is implemented in native Java and uses Repast Simphony software libraries for random number generation and chart creation. Finally, I present several hypothetical scenarios that explore the impact on tax compliance of alternative tax administration proposals.

Friday, May 20 – 3:00 p.m.

Self-organized Criticality in a Model of Mergers and Acquisitions

Benoit Morel, Professor
Department of Engineering and Public Policy
Carnegie Mellon University

Dr. Morel joined the faculty at Carnegie Mellon University in 1987 in the Department of Engineering and Public Policy, with the Program on International Peace and Security. At Carnegie Mellon, his research interests have focused on military high technology, its technical details and structure, and its impact on security and arms control, as well as its effects on American defense policy. Dr. Morel is also interested in non-linear dynamic models, and the study of complex systems and chaos, with application to a variety of areas, such as immunology, fluid mechanics, organization theory, economics, pollution, and environment.

Friday, September 9 – 3:00 p.m.
Research Hall (formerly Research 1)

Title: Differential Adaptive Diffusion: Understanding Diversity and Learning whom to Trust in Viral Marketing

Bill Rand, Asst. Professor
Robert H. Smith School of Business
University of Maryland

Abstract: Viral marketing mechanisms use the existing social network between customers to spread information about products and encourage product adoption. Existing viral marketing models focus on the dynamics of the diffusion process, however they typically: (a) only consider a single product campaign and (b) fail to model the evolution of the social network, as the trust between individuals changes over time, during the course of multiple campaigns. In this work, we propose an adaptive viral marketing model which captures: (1) multiple different product campaigns, (2) the diversity in customer preferences among different product categories, and (3) changing confidence in peers’ recommendations over time. By applying our model to a real-world network extracted from the Digg social news website, we provide insights into the effects of network dynamics on the different products’ adoption. Our experiments show that our proposed model outperforms earlier non-adaptive diffusion models in predicting future product adoptions. We also show how this model can be used to explore new viral marketing strategies that are more successful than classic strategies which ignore the dynamic nature of social networks.

Friday, September 16 – 3:00 p.m.
Center for Social Complexity Suite
Third Floor, Research Hall

Title: Social Networks and Collection Action

William Ferguson, Professor
Department of Economics
Grinnell University

Abstract: Social networks underlie social capital; they are the grid upon or through which all manners of social, economic, and political exchanges occur. We may even think of institutions as networks of information flows and shared understandings, or as complex conduits of cognition, information, and motivation. Networks thus constitute the social foundation for both the existence and resolution of collective action problems (CAPs). Certain CAPs, such as traffic congestion, are fundamentally network problems. For all CAPs, networks influence agents’ understandings of problems and solutions, available information, and their motivation. Network architecture and relationships may underlie CAPs, exacerbate them, and affect possibilities and motivation for resolution. As an underpinning of social capital, social networks can foster trusting relationships—themselves foundations of trade and other exchanges. Moreover, network positions affect the ability of specific individuals, to influence opinions, ideas, and organizational policies: network positions confer power.

Discussion in this chapter is necessarily limited. The field of social networks is both complex and rapidly evolving. Our purpose is to offer essential background with application to political economy and CAPs. Discussion proceeds as follows. Section 1 discusses basic network concepts. Section 2 addresses the impact of localized network structure on the ability of agents to exert influence: network position as a source of power. Section 3 addresses networks as searchable conduits of information. Agents can use local knowledge of network attributes to uncover information about distant individuals and groups. Section 4 discusses information cascades across networks—sequences of rapid adoption of ideas or practices associated with phenomena like fads, financial contagion, or revolution. Section 5 concludes.

Friday, September 23 – 3:00 p.m.
Center for Social Complexity Suite
Third Floor, Research Hall

Title: Complexity and Scale in Networks — From Neurons to Nations

Maksim Tsvetovat, Asst. Professor
Department of Computational Social Science
George Mason University

Abstract: In human history, every culture has produced a set of rules governing social behavior, thus dictating a macro-level social network structure. While these rules do differ from culture to culture, they produce remarkably similar macro-level social networks, both on family- and small-group level, but also on societal level. Examples of
low-level rules can be found in work of Granovetter(1983), diffusion models(Friedkin 1998, Carley 1999, etc), Burt (1992), etc., as well as social-psychological literature such as Simmel(1905) and Heider(1979).

We hypothesize that a basic, low-level set of rules of social behavior may be biologically dictated via the the physiological properties of the human brain. This is based on the findings that most aspects of human (and animal) behavior appear to be at least in part directly mediated by select processes of neurotransmission, including genetically defined individual expression of neurotransmitters and
neuropeptides within distinct brain regions (Schultz 2007). Further, Dunbar(2003) hints at these mechanisms by correlating size of social groups with size of prefrontal cortex in primates.

We call this set of rules a Social Operating System of homo sapiens. In our work, we go down to the hardware to explain social reward mechanisms behind triadic structures, information diffusion, and ultimately social network structure. We are building a first-of-a-kind computational model of this operating system, and will demonstrate
preliminary results

Friday, September 30 – 3:00 p.m.
Center for Social Complexity Suite
Third Floor, Research Hall

Title: A Network Analysis of Global Banking: 1978-2010 (co-authored with Javier Reyes)

Camelia Minoiu, PhD
Economist
IMF Institute
International Monetary Fund

Abstract: We explore the properties of the global banking network using cross-border banking data for 184 countries over 1978–2010. We document an unstable global banking network with empirical distributions of network indicators that change markedly over time, especially for borrowers. Structural breaks in network indicators identify two waves of global capital flows leading up respectively to the 1997–98 East Asian crisis and the 2008–09 global financial crisis. While network density tends to expand and contract with the global cycle of capital flows, network indicators provide incremental information compared to aggregate flows. For example, network density before the global financial crisis was comparable to earlier peaks despite the historically unique build-up in aggregate flows. We also find that country-level financial interconnectedness falls in the aftermath of systemic banking crises and sovereign default episodes. The aftermath of the global financial crisis stands out as an unusual perturbation to the global banking network.