I am a spatial (GIS) agent-based modeler i.e. modeler that simulates the impact of various individual decisions on the environment. My work is mainly methodological i.e. I develop tools that make agent-based modeling (ABM) easier to do. I especially focus on developing tools that allow for evaluating various uncertainties in ABM. One of these uncertainties are the ways of quantifying agent decisions (i.e. the algorithmic representation of agent decision rules) for example to address the question of “How do the agents decide whether to grow crops or rather put land to fallow?”. One of the methods I developed focuses on representing residential developers’ risk perception for example to answer the question: “to what extent is the developer risk-taking and would be willing to build new houses targeted at high-income families (small market but big return on investment)?”. Other ABM uncertainties that I evaluate are various spatial inputs (e.g. different representations of soil erosion, different maps of environmental benefits from land conservation) and various demographics (i.e. are retired farmers more willing to put land to conservation?). The tools I develop are mostly used in (spatial) sensitivity analysis of ABM (quantitative, qualitative, and visual).

Economics, society, complexity, systems, ecosystem, thermodynamics, agent-based modelling, emergent behaviour, evolution.

Global change, human-Earth interactions, complex systems.

The big picture question driving my research is how do complex systems of interactions among individuals / agents result in emergent properties and how do those emergent properties feedback to affect individual / agent decisions. I have explored this big picture question in a number of different contexts including the evolution of cooperation, suburban sprawl, traffic patterns, financial systems, land-use and land-change in urban systems, and most recently social media. For all of these explorations, I employ the tools of complex systems, most importantly agent-based modeling.

My current research focus is on understanding the dynamics of social media, examining how concepts like information, authority, influence and trust diffuse in these new media formats. This allows us to ask questions such as who do users trust to provide them with the information that they want? Which entities have the greatest influence on social media users? How do fads and fashions arise in social media? What happens when time is critical to the diffusion process such as an in a natural disaster? I have employed agent-based modeling, machine learning, geographic information systems, and network analysis to understand and start to answer these questions.

Social network analysis has an especially long tradition in the social science. In recent years, a dramatically increased visibility of SNA, however, is owed to statistical physicists. Among many, Barabasi-Albert model (BA model) has attracted particular attention because of its mathematical properties (i.e., obeying power-law distribution) and its appearance in a diverse range of social phenomena. BA model assumes that nodes with more links (i.e., “popular nodes”) are more likely to be connected when new nodes entered a system. However, significant deviations from BA model have been reported in many social networks. Although numerous variants of BA model are developed, they still share the key assumption that nodes with more links were more likely to be connected. I think this line of research is problematic since it assumes all nodes possess the same preference and overlooks the potential impacts of agent heterogeneity on network formation. When joining a real social network, people are not only driven by instrumental calculation of connecting with the popular, but also motivated by intrinsic affection of joining the like. The impact of this mixed preferential attachment is particularly consequential on formation of social networks. I propose an integrative agent-based model of heterogeneous attachment encompassing both instrumental calculation and intrinsic similarity. Particularly, it emphasizes the way in which agent heterogeneity affects social network formation. This integrative approach can strongly advance our understanding about the formation of various networks.

How to model it: Ecological models, in particular simulation models, often seem to be formulated ad hoc and only poorly analysed. I am therefore interested in strategies and methods for making ecological modelling more coherent and efficient. The ultimate aim is to develop preditive models that provide mechanstic understanding of ecological systems and that are transparent and structurally realistic enough to support environmental decision making.

Pattern-oriented modelling: This is a general strategy of using multiple patterns observed in real systems as multiple criteria for chosing model structure, selecting among alternative submodels, and inversely determining entire sets of unknown model parameters.

Individual-based and agent-based modelling: For many, if not most, ecological questions individual-level aspects can be decisive for explaining system-level behavior. IBM/ABMs allow to represent individual heterogeneity, local interactions, and/or adaptive behaviour

Ecological theory and concepts: I am particularly interested in exploring stability properties like resilience and persistence.

Modelling for ecological applications: Pattern-oriented modelling allows to develop structurally realistic models, which can be used to support decision making and the management of biodiversity and natural resources. Currently, I am involved in the EU project CREAM, where a suite of population models is developed for pesticide risk assessment.

Standards for model communication and formulation: In 2006, we published a general protocol for describing individual- and agent-based models, called the ODD protocol (Overview, Design concepts, details). ODD turned out to be more useful (and needed) than we expected.

Two themes unite my research: a commitment to methodological creativity and innovation as expressed in my work with computational social sciences, and an interest in the political economy of “globalization,” particularly its implications for the ontological claims of international relations theory.

I have demonstrated how the methods of computational social sciences can model bargaining and social choice problems for which traditional game theory has found only indeterminate and multiple equilibria. My June 2008 article in International Studies Quarterly (“Coordination in Large Numbers,” vol. 52, no. 2) illustrates that, contrary to the expectation of collective action theory, large groups may enjoy informational advantages that allow players with incomplete information to solve difficult three-choice coordination games. I extend this analysis in my 2009 paper at the International Studies Association annual convention, in which I apply ideas from evolutionary game theory to model learning processes among players faced with coordination and commitment problems. Currently I am extending this research to include social network theory as a means of modeling explicitly the patterns of interaction in large-n (i.e. greater than two) player coordination and cooperation games. I argue in my paper at the 2009 American Political Science Association annual convention that computational social science—the synthesis of agent-based modeling, social network analysis and evolutionary game theory—empowers scholars to analyze a broad range of previously indeterminate bargaining problems. I also argue this synthesis gives researchers purchase on two of the central debates in international political economy scholarship. By modeling explicitly processes of preference formation, computational social science moves beyond the rational actor model and endogenizes the processes of learning that constructivists have identified as essential to understanding change in the international system. This focus on the micro foundations of international political economy in turn allows researchers to understand how social structural features emerge and constrain actor choices. Computational social science thus allows IPE to formalize and generalize our understandings of mutual constitution and systemic change, an observation that explains the paradoxical interest of constructivists like Ian Lustick and Matthew Hoffmann in the formal methods of computational social science. Currently I am writing a manuscript that develops these ideas and applies them to several challenges of globalization: developing institutions to manage common pool resources; reforming capital adequacy standards for banks; and understanding cascading failures in global networks.

While computational social science increasingly informs my research, I have also contributed to debates about the epistemological claims of computational social science. My chapter with James N. Rosenau in Complexity in World Politics (ed. by Neil E. Harrison, SUNY Press 2006) argues that agent-based modeling suffers from underdeveloped and hidden epistemological and ontological commitments. On a more light-hearted note, my article in PS: Political Science and Politics (“Clocks, Not Dartboards,” vol. 39, no. 3, July 2006) discusses problems with pseudo-random number generators and illustrates how they can surprise unsuspecting teachers and researchers.