Agent-Based Computational Sociology

Flaminio Squazzoni

• Publisher: Wiley
• Year: 2012
• Print ISBN: 9780470711743
• eBook ISBN: 9781119941637
• Edition: 1
• Topic: Social Sciences
• Subtopic: Social Science Research & Methodology
• Index: Social Sciences

Most of the intriguing social phenomena of our time, such as international terrorism, social inequality, and urban ethnic segregation, are consequences of complex forms of agent interaction that are difficult to observe methodically and experimentally. This book looks at a new research stream that makes use of advanced computer simulation modelling techniques to spotlight agent interaction that allows us to explain the emergence of social patterns. It presents a method to pursue analytical sociology investigations that look at relevant social mechanisms in various empirical situations, such as markets, urban cities, and organisations.

This book:

• Provides a comprehensive introduction to epistemological, theoretical and methodological features of agent-based modelling in sociology through various discussions and examples.
• Presents the pros and cons of using agent-based models in sociology.
• Explores agent-based models in combining quantitative and qualitative aspects, and micro- and macro levels of analysis.
• Looks at how to pose an agent-based research question, identifying the model building blocks, and how to validate simulation results.
• Features examples of agent-based models that look at crucial sociology issues.
• Supported by an accompanying website featuring data sets and code for the models included in the book.

Agent-Based Computational Sociology is written in a common sociological language and features examples of models that look at all the traditional explanatory challenges of sociology. Researchers and graduate students involved in the field of agent-based modelling and computer simulation in areas such as social sciences, cognitive sciences and computer sciences will benefit from this book.

1

What is agent-based computational sociology all about?

There is no doubt that the last twenty years have brought radical changes to the use of computers in social research (Heise and Simmons 1985; Gilbert and Abbott 2005). In the past (and even still today), social scientists used computers to provide analytic solutions to complicated equation systems that represented a given system's structure, or more generally to estimate statistical models for data. From the 1990s onward, they started to use advanced computational techniques in an innovative way to simulate and analyze implications of agent interaction in social structures (e.g., Epstein and Axtell 1996; Axelrod 1997a; Epstein 2006; Miller and Page 2007).

Computational sociology, that is, the use of computationally intensive methods to model social phenomena, is not a recent development (Brainbridge 2007). It is a branch of sociology that has a long and, to a certain extent, venerable tradition that goes back to the 1960s. At that time, under the influence of systems theory and structural functionalism, computer simulation was used to model control and feedback mechanisms in systems, such as organizations, cities, or global populations. The idea was to simulate complicated differential equation models to predict population distribution as a function of systemic factors, such as urban traffic, migration, demographic change, or disease transmission. Inspired by Forrester's work on world dynamics (Forrester 1971) and the idea of systems theory and cybernetics, the focus was on systems and aggregates rather than on agents and behavior, and on prediction rather than understanding and explanation (Sawyer 2005).

Nevertheless, against this trend, some pioneers started to use computer simulation to investigate models of micro social processes. In the 1960s, James S. Coleman led the most active research center for computer research in sociology in the US. At Johns Hopkins University, he published some interesting contributions to plead the cause of simulation models in sociology aiming to investigate agent interaction (Coleman 1962, 1964b). Raymond Boudon also published an article on prosecutions in France with a simulation model as a key element (Davidovitch and Boudon 1964). Some years later, in his famous book on the mathematical approach to sociology, he systematically examined the similarities and differences of equation-based and computer simulation models to understand social processes from a micro–macro perspective (Boudon 1970). Ahead of their times, these leading sociologists pre-empted the agent turn of the 1990s.

Computer simulation approaches and techniques changed over time, as we will see later. From the 1990s onward, sociologists started to analyze macro social aggregates as the resultant properties of micro interaction, by explicitly modeling agents, interaction and environment (i.e., geographical space, institutional settings, and/or social structures). The growth of computational capacity applied to research, as well as its ubiquitous and distributed nature, allowed the creation and diffusion of the first agent-based model (ABM) open-source simulation platforms, easily manageable even with portable computers. This innovation in research technologies set the stage for an ‘agent-based turn’ in social research and helped agent-based computational sociology to materialize.

The aim of this chapter is to introduce agent-based computational sociology as the study of social patterns by computer models of social interaction between heterogeneous agents embedded in social structures (e.g., social networks, spatial neighborhoods, or institutional scaffolds). The first section identifies predecessors and founding fathers. Herbert Simon, James S. Coleman, Raymond Boudon, Thomas Schelling and Mark Granovetter have been included as they pioneered and/or largely influenced this type of research. As we will see, not only is there a certain coherence between the work of these authors, but also a fil rouge links these studies to certain streams of sociology today that revolve around the idea of the ‘generative’ approach to sociological investigation. These research streams have recently been systematized under the name of ‘analytical sociology’, where emphasis is given to the explanation of social patterns from agent interaction and where agent-based modeling plays a pivotal role (Hedström and Swedberg 1998; Hedström 2005; Bearman and Hedström 2009; Hedström and Ylikoski 2010; Manzo 2010).

The second section illustrates the main ideas of this new type of sociology. They are as follows: (a) the primacy of models, (b) the generative approach to explanation, (c) a pragmatic approach to the micro–macro link, (d) process and change as key elements of sociological investigation, (e) a reconciliation of deduction and induction, theory and data through models, and (f) a tendency towards a trans-disciplinary/issue-oriented style of research. While ideas from (a) to (e) have application in the literature, (f) is still in the latent phase, though not less important.

The third section illustrates ABMs as the tool that has made this new type of research possible. It is worth noting that agent-based computational sociology does not totally conflate with ABMs, as the latter is used also in physics, biology and computer sciences with different purposes. Here, ABMs specifically target the properties of social behavior and interaction and addresses relevant empirical sociological puzzles. Without entering into technical detail, a comparison between ABMs and other simulation techniques is presented that examines the peculiarities of ABMs for sociological investigation. Finally, the fourth section looks at ABM classification, illustrating differences in its use in research. Certain implications are made to link theory and empirical data treated in detail in Chapter 4. Examples and attention to substantive issues will be looked at in Chapters 2 and 3.

1.1 Predecessors and fathers

Perhaps unexpectedly, the ABM approach has a venerable legacy in sociology. In common with traditional mathematical sociology, it includes the idea that formalized models can make sociology more scientific (Coleman 1964a; Fararo 1969). However, it makes no sacrifice to analytic solutions or top-down deductions. Indeed, the ABM perspective espouses a more complex view of sociological models, where (a) theory should be the result of bottom-up data exploration, and (b) models should look at nonlinear local agent interaction and global out-of-equilibrium system behavior, rather than pre-constituted structural behavior and equilibrium. It shares the idea that computational formalization can help to improve the theory building process, by revealing non-obvious mechanisms and providing for a theory test.

However, the ABM approach aims to develop sociologically rich models that mimic the properties of social behavior and interaction. Indeed, ABMs aim to understand what is taken for granted in more functionalistic, macro-oriented simulation approaches, such as system dynamics, that is, the emergence of social patterns, structures and behavior from agent interaction. This also helps us to understand under which conditions, certain social patterns might emerge in reality.

One of those who has contributed the most to this type of research, is a non-sociologist, the Nobel Prize winner, Herbert A. Simon. One of the most prominent social scientists of the last century, Simon influenced a wide range of disciplines, from artificial intelligence to organization science and psychology. One of his simplest ideas was that there is no isomorphism between the complexity that social systems show at the macro level and their complexity at the micro level. In many cases, the former is nothing more than the result of interaction between simple micro processes. Therefore, computer simulation is pivotal to simplify and model complex social systems from a micro–macro approach (e.g., Simon 1969).

Simon was also interested in understanding what he called ‘poorly understood systems’, that is, those in which the modeler has poor or no knowledge of the laws that govern inner systems. By suggesting the rationale for simple computer simulation models that look at these types of systems, he argued that:

resemblance in behaviour of systems without identity of the inner systems is particularly feasible if the aspects in which we are interested arise out of the organization [italics in original] of the parts, independently of all but a few properties of the individual components (Simon 1969, p. 17).

The first lesson is that understanding the interaction mechanisms between individual components is pivotal to look at the complex behavior of a social system. The second is that by pinpointing the explanatory power of interaction, sociology can omit detailed knowledge of the behavior of each individual component, at the same time avoiding referring to supposed causal autonomy of macro social entities to understand macro system behavior. The latter should be viewed as fully shaped by organizational micro processes.

Obviously, this was not Simon's only contribution to the ABM approach in social sciences. We can mention his investigation into the foundations of human behavior and his theory on bounded rationality (Simon 1982). He influenced all ABM scientists who have tried to understand how a population of boundedly rational agents can spontaneously and endogenously give rise to patterns of collective intelligence in diverse spheres of the economy and society (e.g., Epstein and Axtell 1996).1

The Nobel Prize winner Thomas C. Schelling, with his pioneering work on the micro–macro link and his famous segregation model in the 1970s, has had an incommensurable influence on agent-based computational sociology. In the first pages of his influential book Micromotives and Macrobehavior in 1978, he illustrated the crucial challenge of understanding macro behavior from agent interaction by using this simple example:

There are easy cases, of course, in which the aggregate is merely an extrapolation from the individual. If we know that every driver, on his own, turns his lights on at sundown, we can guess that from our helicopter we shall see all the car lights in a local area going on at about the same time. […] But if most people turn their lights on when some fraction of the oncoming cars already have their lights on, we'll get a different picture from our helicopter. In the second case, drivers are responding to each others’ behaviour and influencing each other's behaviour. People are responding to an environment that consists of other people responding to their [italics in original] environment, which consists of people responding to an environment of people's responses. Sometimes the dynamics are sequential […]. Sometimes the dynamics are reciprocal […]. These situations, in which people's behaviour or people's choices depend on the behaviour or the choices of other people, are the ones that usually don't permit any simple summation or extrapolation to the aggregates. To make that connection we usually have to look at the system of interaction [italics in original] between individuals and their environment […]. And sometimes the results are surprising. Something they are not easily guessed. Sometimes the analysis is difficult. Sometimes it is inconclusive. But even inconclusive analysis can warn against jumping to conclusions about individual intentions from observations of aggregates, or jumping to conclusions about the behaviour of aggregates from what one knows or can guess about individual intentions (Schelling 1978, pp. 13–14).

The difficulty of mapping micro and macro levels when nonlinear agent interactions are involved is the premise that avoids both conflating or contrasting the various levels of social system analysis (Squazzoni 2008). This was evident in Schelling's famous segregation model, which is now a standard example in ABM literature, where the dynamics of residential mobility and segregation by race and ethnicity, that is, a long lasting pattern of many large cities in the US, were explained not as the result of racist preferences, but...

Table of contents

• Cover
• Title Page
• Copyright
• Dedication
• Preface
• Chapter 1: What is agent-based computational sociology all about?
• Chapter 2: Cooperation, coordination and social norms
• Chapter 3: Social influence
• Chapter 4: The methodology
• Chapter 5: Conclusions
• Appendix A
• Appendix B
• Index