Oslo Summer School in Comparative Social Science Studies 2018

Introduction to Agent-based Modeling and Computing

Professor Robert Axtell, Department of Computational and Data Sciences, College of Science, Department of Economics, College of Humanities and Social Sciences Center for Social Complexity, Krasnow Institute for Advanced Study, George Mason University, USA

Course dates: 23 - 27 July 2018

Download PDF

Main disciplines: Economics, Sociology, Data Science
Political Science, Research Methodology

Course Credits: 10 pts (ECTS)
Limitation: 25 participants

Course objectives / learning outcome
This course is designed to introduce social scientists (economists, political scientists, sociologists, anthropologists, finance students, linguists, etc.) to the emerging paradigm of agent-based modeling (ABM), in which each individual in a social system is represented as a distinct software object.

These objects, called agents, have internal data (states) and methods (rules of behavior) and interact directly with one another according to behavioral rules. From specific initial conditions such models move forward in time through agent interactions and social structures emerge, from market-clearing prices to behavioral norms and social conventions. This ‘bottom-up’ perspective on social phenomena contrasts with so-called ‘representative agent’ formulations common in economics and are a principled way to study collective behavior in sociology, for instance. Agents in ABMs can be arbitrarily heterogeneous in contrast to game theory where typically only a few agent ‘types’ are permitted. Agents typically have heterogeneous information and are boundedly rational. Since agent interactions generally give rise to social networks there is a definite sense in which ABM generalizes social network theory, permitting the study of truly dynamic networks.

Furthermore, while it is common for ABMs to yield steady-states at the aggregate, social level, these are often accompanied by perpetual flux and novelty at the agent level, meaning Nash equilibria and other ‘solution concepts’ from conventional game theory are of limited relevance. We will illustrate all of the above through a series of progressively more complex models, some of which are grounded in micro-data.

Students will learn how to manipulate existing ABMs as well as how to program such models from scratch. While some familiarity with computer programming will be helpful for this course, a user-friendly software package, NetLogo, will be used extensively and students new to programming usually find it easy to learn.

Some knowledge of probability and statistics is helpful, but no advanced mathematics is required. Concerning programming, it is preferable if students can program in some language or another, but this is not a hard requirement. We will be using a public domain software package and one path through the course material will be exercising existing models while the more standard path will be programming models. Really, as long as a student has some basic computational background, such as spreadsheet skills, then they should be able to follow along.


Session 1: Agent-based modeling (ABM) in the social sciences
We will cover the main rationales for employing ABM, including (1) relaxing rational actor specifications, (2) adding networks to mathematical models that do not have them, (3) investigating interactions away from equilibrium (e.g., Nash equilibria), (4) the introduction of agent heterogeneity, and (5) understanding emergent social phenomena, like traffic jams, fads, market bubbles, and so on.

Relevant literature includes:

  1. R. L. Axtell, in Proceedings of the Workshop on Agent Simulation: Applications, Models, and Tools, C. M. Macal, D. Sallach, Eds. (Argonne National Laboratory, Chicago, Illinois, 2000), pp. 3-24.
  2. R. L. Axtell, What economic agents do: How cognition and interaction lead to emergence and complexity. Review of Austrian Economics 20, 105 (2007).
  3. N. Gilbert, Agent-Based Models. Quantitative Applications in the Social Sciences book series, J. Fox, ed. (Sage Publications, Inc., Thousand Oaks, CA, 2008)

Session 2: Software for agent-based modeling
We will introduce some of the main software packages and frameworks for ABM, including RePast, MASON, and NetLogo, along with the pros and cons of writing lower level code in C/C++, Java, Python, and other languages. Students who will use a language to write ABMs in this course should have experience programming in that language and have access to appropriate manuals. Students who know Java may find RePast or MASON useful. For those students who are new to programming instruction will be provided for using NetLogo, and one of the following book may prove useful as a reference:

  1. S. F. Railsback, V. Grimm, Agent-Based and Individual-Based Modeling: A Practical Introduction.  (Princeton University Press, Princeton, N.J., 2011).
  2. U. Wilensky, W. Rand, An Introduction to Agent-Based Modeling: Modeling Natural, Social, and Engineered Complex Systems with NetLogo.  (MIT Press, Cambridge, Mass., 2015).

Session 3: A simple agent-based model: zero-intelligence traders
During this meeting we will build a simple ABM of trading (market) behavior. It is based on the following paper but those who are not economists may find the paper a bit technical. Students of economics may find the second paper interesting.

  1. D. K. Gode, S. Sunder, Allocative Efficiency of Markets with Zero-Intelligence Traders: Market as a Partial Substitute for Individual Rationality. Journal of Political Economy CI, 119 (1993).
  2. D. Cliff, J. Bruten, “Minimal-Intelligence Agents for Bargaining Behaivors in Market-Based Environments” HPL-97-91 (Hewlett-Packard Labs, Bristol, UK, 1997).

Session 4: Applications of ABMs in economics, finance, game theory, and related disciplines
We will survey various uses of agent computing in these fields, with the content somewhat tailored to student interest. An excellent source of materials in this broad area is Professor Leigh Tesfatsion’s website dedicated to agent-based computational economics (ACE):

  • http://www2.econ.iastate.edu/tesfatsi/ace.htm

Session 5: Agents create social networks and applications of ABMs in sociology
The uses of ABM in sociology will be reviewed. One tenet of the ‘analytical sociology’ movement is to provide micro-specifications of agent actions, which we always do with ABM, so that emerging area is quite relevant.

  1. M. W. Macy, R. Willer, From Factors to Actors: Computational Sociology and Agent-Based Modeling. Annual Review of Sociology 28, 143 (2002).
  2. P. Hedstrom, Dissecting the Social: On the Principles of Analytical Sociology. (Cambridge University Press, New York, N.Y., 2005).

Session 6: Agent heterogeneity, big data in the social sciences, and synthetic populations
Building empirically-grounded ABMs involve specification of individual people and their behavior. The application of machine learning and other techniques from data science for the creation of ABMs is a topic at the research frontier. We will survey this active area.

  1. Adiga, A., A. Agashe, S. Arifuzzaman, C. L. Barrett, R. Beckman, K. Bisset, J. Chen, Y. Chungbaek, S. Eubank, S. Gupta, M. Khan, C. J. Kuhlman, E. Lofgren, B. Lewis, A. Marathe, M. V. Marathe, H. S. Mortveit, E. Nordberg, C. Rivers, P. Stretz, S. Swarup, A. Wilson and D. Xie (2015). Generating a synthetic population of the United States. Network Dynamics and Simulation Science Laboratory Technical Report, Virginia Tech.
  2. Lamperti, F., A. Roventini and A. Sani (2017). Agent-Based Model Calibration using Machine Learning Surrogates. Available online.

Session 7: Specifying agent behavior: incorporating laboratory data into ABMs
Use of data from experiments to calibrate ABMs is another active topic at the research frontier. We will survey this broad area.

1. Hommes, C. H. (2011). "The Heterogeneous Expectations Hypothesis: Some Evidence From the Lab." Journal of Economic Dynamics and Control 35(1): 1-24.

Session 8: Large-scale models (millions of agents or more)
Modeling millions of agents or more is a growing area of application of ABMs. We will look at some of these models, including

  1. Axtell, R. L. (2016). 120 Million Agents Self-Organize into 6 Million Firms: A Model of the U.S. Private Sector. Proceedings of the 15th International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS2016). J. Thangarah, K. Tuyls, C. Jonker and S. Marsella. Singapore, International Foundation for Autonomous Agents and Multiagent Systems.

Session 9: Calibration, estimation, verification, and validation of ABMs
There are many approaches to making ABMs empirically-relevant. We will survey some of these, including:

  1. Axtell, R. L. and J. M. Epstein (1994). "Agent-Based Models: Understanding Our Creations." Bulletin of the Santa Fe Institute.
  2. Axtell, R. L., R. Axelrod, J. M. Epstein and M. D. Cohen (1996). "Aligning Simulation Models: A Case Study and Results." Computational and Mathematical Organization Theory 1(2): 123-141.
  3. Alfarano, S., T. Lux and F. Wagner (2005). "Estimation of Agent-Based Models: The Case of an Asymmetric Herding Model." Computational Economics 26(1): 19-49.

Session 10: Advanced topics, based on student interest: agent activation regimes, applications in anthropology/archaeology, relation of ABMs to cellular automata and interacting particle systems, macroeconomic modeling using ABMs, applications of ABMs in politics, computational statistics and ABMs, commercial applications of ABMS, or complex systems and ABMs.

Some relevant references include:

  1. R. L. Axtell et al., Population Growth and Collapse in a Multiagent Model of the Kayenta Anasazi in Long House Valley. Proc Natl Acad Sci U S A 99, 7275 (May 14, 2002).
  2. M. Laver, E. Sergenti, Party Competition: An Agent-Based Model. (Princeton University Press, Princeton, NJ, 2011).
  3. V. Darley, A. V. Outkin, A NASDAQ Market Simulation: Insights on a Major Market from the Science of Complex Adaptive Systems. N. Johnson, Ed., Complex Systems and Interdisciplinary Science (World Scientific, 2007), vol. 1.

The lecturer
Professor Robert Axtell earned an interdisciplinary Ph.D. degree at Carnegie Mellon University, where he studied computing, social science, and public policy. His teaching and research involves computational and mathematical modeling of social and economic processes. Specifically, he works at the intersection of multi-agent systems computer science and the social sciences, building so-called agent-based models for a variety of market and non-market phenomena.

His research has been published in the leading scientific journals, including Science and the Proceedings of the National Academy of Sciences, USA, and reprised in Nature, and has appeared in top disciplinary journals (e.g., American Economic Review, Computational and Mathematical Organization Theory, Economic Journal), in general interest journals (e.g., PLOS One) and in specialty journals (e.g., Journal of Regulatory Economics, Technology Forecasting and Social Change.)

Tags: Sociology, Economics, PhD, Data Science, Summer School, Political Science, Agent-based Modeling
Published Nov. 29, 2017 9:02 AM - Last modified Aug. 30, 2018 9:02 AM