Agent-Based Modeling

[oxbibsubmitdefin.docx] 1 Agent-Based Modeling Edmund Chattoe-Brown, Department of Sociology, University of Leicester Agent-Based Modeling Introduction General Overviews and Textbooks Reviews and Reference Works The Methodology of Agent-Based Modeling Uses of Agent-Based Modeling Analyzing Agent-Based Models Cognitive and Environmental Elements of Agent-Based Models Interactional and Social Elements of Agent-Based Models Data and Agent-Based Models Institutions of Agent-Based Modeling Challenges to Agent-Based Modeling Introduction Agent-Based Modeling is a research method that represents theories of social behavior as computer programs of a particular kind, rather than narratives (as ethnography does) or equations (as regression models do). Like existing research methods in sociology (both qualitative and quantitative) it can be applied throughout the discipline and offers advantages for certain research questions. In particular, the approach is referred to as agent-based because the computer program unambiguously represents interactions between heterogeneous social actors while also explicitly determining their aggregate simulated consequences. This distinguishes Agent-Based Modeling from existing quantitative approaches in sociology where the relationship between aggregate associations and individual agency is often unclear. It also distinguishes the method from existing qualitative approaches that, while investigating individuals and their interactions, have no systematic techniques for establishing their aggregate consequences. Given this capability, the methodology of Agent-Based Modeling has a distinctive logic. Agent-Based Models are calibrated using data on individual behavior (for example using ethnography or laboratory experiments) and then the computer program generates simulated aggregate data. This can then be compared with equivalent real data for validation. It is the independence of these two activities that provides Agent-Based Modeling with its distinctive claim to explanatory power. The explicitly represented link between individual and aggregate respects the complexity of social systems, the phenomenon in which individuals and their simple interactions may produce surprisingly counter-intuitive aggregate outcomes. Agent-Based Models are thus particularly suitable for investigating sociological issues involving heterogeneous actors, diverse cognitive processes and social systems mediated by entities operating between the level of the individual and the aggregate (like schools and churches). [oxbibsubmitdefin.docx] 2 One aim of this bibliography is to strike a balance between technical aspects of the method (available programming languages, calibration and validation) and important or distinctive applications in diverse areas of sociological interest. Another is to stress the importance of the distinctive methodology in maintaining (as with existing methods) the scientific quality of research. General Overviews and Textbooks As a relatively newly established field, the number of overviews and textbooks is not overwhelming. Epstein and Axtell 1996 is the earliest book length introduction but one can only consider it as a textbook with difficulty. For a number of years, Gilbert and Troitzsch 2005 was the sole textbook available (though they also dealt with social science simulation more broadly). Gilbert 2008 has now produced an excellent short introduction dealing specifically with Agent-Based Modeling. More recently there has been a significant increase in the publication rate somewhat ramified by the range of different programming languages now available for Agent-Based Modeling. There is something of a tradeoff between books organized around particular programming languages (which tend to emphasize technical matters and choose examples to suit) and those organized around particular domains or disciplines (which tend to emphasize substantive issues). Railsback and Grimm 2012 and Wilensky and Rand 2015, both based on the very accessible programming language Netlogo, fall into the former class while Squazzoni 2012 (dealing with core sociological topics like cooperation) and Miller and Page 2007 (with their emphasis on the social science implications of complexity) fall into the latter. Epstein, Joshua and Robert Axtell. 1996. Growing Artificial Societies: Social Science from the Bottom Up. Washington, DC: Brookings Institution Press. Organized around examples of basic social processes in the imaginary Sugarscape this book has limitations despite its importance. The program code was not originally available (though now reimplemented by others) and the examples sometimes emphasize individualistic simplicity rather than social plausibility. Gilbert, Nigel. 2008. Agent-Based Models. Thousand Oaks, CA: Sage. Extremely concise and based around a sociologically imaginative example of subjective group definition and membership. Gilbert, Nigel and Klaus Troitzsch. 2005. Simulation for the Social Scientist, 2nd ed. Maidenhead: Open University Press. This is still a very useful and accessible introduction to Agent-Based Modeling. Although some chapters cover different simulation approaches, others (like that implementing evolutionary approaches to social change) retain wider relevance to sociology. [oxbibsubmitdefin.docx] 3 Miller, John and Scott Page. 2007. Complex Adaptive Systems: An Introduction to Computational Models of Social Life. Princeton, NJ: Princeton University Press. The strength of this book is its focus on the relevance of complexity and emergence for social science understanding. The examples are somewhat more stylized and tend to impart inspiration rather than technical skills. Railsback, Steven and Volker Grimm. 2012. Agent-Based and Individual-Based Modeling: A Practical Introduction. Princeton, NJ: Princeton University Press. This book teaches Netlogo programming and ABM methodology very effectively through staged examples but only some of these will be relevant to sociologists. Squazzoni, Flaminio. 2012. Agent-Based Computational Sociology. Chichester, UK: Wiley. A good all round introduction, particularly useful for its specific sociological focus, its unusual use of experimental data and extensive reference and resource sections. Wilensky, Uri and William Rand. 2015. An Introduction to Agent-Based Modeling: Modeling Natural, Social, and Engineered Complex Systems with NETLogo. Cambridge, MA: The M. I. T. Press. Written by developers of the programming language, this is a technically excellent textbook but for sociologists quite a few examples are less immediately relevant. Reviews and Reference Works There is something of a tension in the history of Agent-Based Modeling. On one hand, publications date quite rapidly from a technical point of view. On the other, as a relatively specialized field with some obscure sources, past research has not always been effectively incorporated into existing practice. I have thus struck a balance between recent material (which is better technically) and older, sometimes neglected, contributions (whose full intellectual value has not yet been exhausted). Under the first heading Halpin 1999 is a useful starting point for the older literature in sociology. By contrast, Bianchi and Squazzoni 2015 have the advantage of being extremely recent. Another very useful review article with slightly different emphases (and a broader remit) is Squazzoni, et al. 2014. Under the second heading, to concisely illustrate my point, none of the three examples above cite Dutton and Starbuck 1971, an early and substantial collection of work including sociological applications and (more surprisingly given its high status journal) Gullahorn and Gullahorn 1965 which is particularly important given its early date. Finally, this is an appropriate section in which to indicate an effective and compendious anthology (Gilbert 2010), a relatively recent handbook (Edmonds and Mayer 2013) and a typical example of the substantial edited volumes common in the field during an era when journal outlets remained quite rare (Suleiman, et al. 2000). [oxbibsubmitdefin.docx] 4 Bianchi, Federico and Flaminio Squazzoni. 2015. Agent-based models in sociology. Wiley Interdisciplinary Reviews: Computational Statistics 7: 284-306. [doi:10.1002/wics.1356] A recent and very densely packed review though sociologists might not entirely share the characterization of sociology that the chosen examples suggest. Dutton, John and William Starbuck, eds. 1971. Computer simulation of human behavior. New York, NY: Wiley. This was an important and exhaustive contribution in its time. Many of the issues it raises about data and methodology remain challenging even today. Edmonds, Bruce and Ruth Meyer, eds. 2013. Simulating social complexity. Berlin: Springer. An extremely thorough and detailed attempt to combine advanced overviews with practical guidance for prospective users of Agent-Based Modeling. Gilbert, Nigel, ed. 2010. Computational social science, 4 vols. London: Sage. A very useful collection of important articles selected in collaboration with the Agent-Based Modeling community, some of which are relatively inaccessible otherwise. Gullahorn, John and Jeanne Gullahorn. 1965. Some Computer applications in social science. American Sociological Review 30: 353-365. An important early overview providing otherwise neglected references and a detailed example paying careful (if flawed) attention to calibration and validation issues. Halpin, Brendan. 1999. Simulation in sociology. American Behavioral Scientist 42: 1488-1508. [doi:10.1177/0002764299042010003] Although covering a wider range of simulation approaches, this review gives a clear and balanced overview of relevant research up to that time. Squazzoni, Flaminio, Wander Jager, and Bruce Edmonds. 2014. Social simulation in the social sciences: A brief overview. Social Science Computer Review 32: 279-294. [doi:10.1177/0894439313512975] The organization of this review article, around different levels of description (micro, meso and macro) gives it a novel emphasis. Suleiman, Ramzi, Klaus Troitzsch, and Nigel Gilbert, eds. 2000. Tools and Techniques for Social Science Simulation. Heidelberg: Physica. Although some of the technical material has now dated, the substantive concerns and range of topics give an effective sense of the field’s liveliness. [oxbibsubmitdefin.docx] 5 The Methodology of Agent-Based Modeling The status of Agent-Based Modeling methodology is somewhat unexpected. Although everyone seems to agree what its distinctive advantages are (and although I can find no developed alternatives in the literature to suggest a controversy) it is rather rarely followed in practice. The methodology is most concisely stated in Gilbert and Troitzsch 2005, but Epstein 2007 provides an extended discussion. Ostrom 1988 is important in articulating the key point that representations involving computer programs can be substantively different from those using equations and narratives. Based on an example dealing with ethnic residential segregation by Schelling 1971 (which has achieved classic status despite being typically neither calibrated nor validated) Chattoe-Brown 2013 explores the distinctiveness of Agent-Based Modeling and some implications for sociological practice in more depth. In a surprisingly neglected article, Hägerstrand 1965 provided an early demonstration of the methodology. Further excellent examples of the feasibility of calibration and validation continue to be forthcoming (Abdou and Gilbert 2009) but seem to have made surprisingly little difference to general practice so far, a point rigorously documented by Angus and Hassani-Mahmooei 2015. Abdou, Mohamed and Nigel Gilbert. 2009. Modelling the emergence and dynamics of social and workplace segregation. Mind and Society 8: 173-191. [doi:10.1007/s11299-009-0056-3] This article provides a carefully argued example illustrating the feasibility of the calibration and validation approach with impressive results. Angus, Simon and Behrooz Hassani-Mahmooei. 2015. “Anarchy” reigns: A quantitative analysis of agent-based modelling publication practices in JASSS, 2001-2012. Journal of Artificial Societies and Social Simulation 18: [http://jasss.soc.surrey.ac.uk/18/4/16.html] Although mainly dealing with the effective presentation of results, this article also quantifies the dominance of Agent-Based Models lacking calibration and validation. Chattoe-Brown, Edmund. 2013. Why sociology should use agent based modelling. Sociological Research Online, 18: [http://www.socresonline.org.uk/18/3/3.html] This article illustrates the significance of Agent-Based Modeling and its methodology for sociological practice focusing on the integration of qualitative and quantitative data, the interpretation of social causation and the challenge of rigorous theory building. Epstein, Joshua 2007. Generative Social Science: Studies in Agent-Based Computational Modeling. Princeton, NJ: Princeton University Press. A detailed discussion of the generativist manifesto is provided in chapter 1. Gilbert, Nigel and Klaus Troitzsch. 2005. Simulation for the Social Scientist, 2nd ed. Maidenhead: Open University Press. [oxbibsubmitdefin.docx] 6 Chapter 2 provides a very concise overview of the methodology with an insightful comparison to statistical modeling. Hägerstrand, Torsten. 1965. A Monte Carlo approach to diffusion. European Journal of Sociology 6: 43-67. [doi:10.1017/S0003975600001132] This is one of the earliest Agent-Based Models combining calibration and validation with a clearly sociological application area. Ostrom, Thomas 1988. Computer simulation: The third symbol system. Journal of Experimental Social Psychology 24: 381-392. [doi:10.1016/0022-1031(88)90027-3] This early article is programmatically important and, in articulating the logic of the third symbol system, remains highly topical. Schelling, Thomas 1971. Dynamic models of segregation. Journal of Mathematical Sociology 1: 143186. [doi:10.1080/0022250X.1971.9989794] An early example dealing with an existing sociological issue (ethnic residential segregation) and showing how simple individual actions and interactions can lead to surprising aggregate effects. Uses of Agent-Based Modeling Although Agent-Based Models can be used in several different ways, the fact that the methodology is rarely followed creates problems in this area, namely that it is often impossible to tell what standards of evaluation authors intend should apply to their work. The danger is that there are too many AgentBased Models (with contradictory – or at least very divergent – assumptions) that are merely not implausible. Takács and Squazzoni 2015 provide a useful example of the thought experiment use for Agent-Based Modeling in showing how it is possible for labor market inequality to occur without discrimination or differences in quality across employable groups. (Even thought experiments may be important where some arbitrary models – those of policy-makers that are backed by power for example – can have real consequences.) Hamill and Gilbert 2009 illustrate the instrumental use of Agent-Based Modeling. Their model is not intended to describe social network formation but in generating empirically plausible network structures it is nonetheless valuable in other substantive applications. Chattoe 2006 demonstrates the theory evaluation use of Agent-Based Modeling (another area where arbitrary models, this time among sociologists, can have real effects) by showing that an explanation of the “strict churches are strong” result proposed by Iannaccone fails under more realistic assumptions in which church formation and failure are dynamic. The distinctive mixture of formality and richness in Agent-Based Modeling also provides a theory development use in areas where sociology has struggled (or less constructively abstained) till now. Good examples of this use are studies of adaptive behavior (Macy 1990), social structure (Lomborg 1996) and social differentiation (Mark 1998). A good example of the policy exploration use is provided by Scott, et al. 2016 who simulate the social harms (and possible interventions) as increasingly drunken agents stagger round an artificial Melbourne. (A properly calibrated and validated Agent-Based Model is much cheaper and [oxbibsubmitdefin.docx] 7 safer to experiment with than reality.) Janssen and Jager 1999 provide an early illustration of the theory synthesis use of Agent-Based Models by exploring the effects of different claims about the nature of decision processes (rational, habitual, adaptive) that can be found across the social sciences. Finally, as might be expected from a general research method, mixed approaches are also possible. Bravo, et al. 2012 contribute to research on the evolution of social networks using experiments as data but then develop an Agent-Based Model to study outcomes in networks larger than would be experimentally attainable. Bravo, Giangiacomo, Flaminio Squazzoni, and Riccardo Boero. 2012. Trust and partner selection in social networks: An experimentally grounded model. Social Networks 34: 481-492. [doi:10.1016/j.socnet.2012.03.001] This example is relatively unusual in using laboratory experiments as data but also provides an interesting example of a mixed methods approach to Agent-Based Modeling. Chattoe, Edmund. 2006. Using simulation to develop testable functionalist explanations: A case study of church survival. The British Journal of Sociology 57: 379-397. [doi:10.1111/j.1468- 4446.2006.00116.x] This article uses Agent-Based Modeling to contribute both specifically and generally to sociological theory. It shows how Iannaccone’s account of church strictness fails under less restrictive assumptions but also implements a coherent functionalist account of church survival. Hamill, Lynne and Nigel Gilbert. 2009. Social circles: A simple structure for agent-based social network models. Journal of Artificial Societies and Social Simulation 12: [http://jasss.soc.surrey.ac.uk/12/2/3.html] This is an example of an Agent-Based Model that does a practical – or engineering – task (generating empirically plausible network structures) rather than describing a target social system per se. Janssen, Marco and Wander Jager. 1999. An integrated approach to simulating behavioural processes: A case study of the lock-in of consumption patterns. Journal of Artificial Societies and Social Simulation 2: [http://jasss.soc.surrey.ac.uk/2/2/2.html] This article shows how approaches from different disciplines can be integrated using AgentBased Modeling. This counters the problem that psychology, economics and sociology take for granted potentially incompatible views about the nature of decision-making. Lomborg, Bjørn. 1996. Nucleus and shield: The evolution of social structure in the iterated Prisoner’s Dilemma. American Sociological Review 61: 278-307. This article uses an Agent-Based Model to formalize ideas of social evolution and heterogeneous decision that illuminate the spontaneous emergence of social structure. [oxbibsubmitdefin.docx] 8 Macy, Michael 1990. Learning theory and the logic of critical mass. American Sociological Review 55: 809-826. This article uses an Agent-Based Model to formalize the intuition that adaptation can endogenously sustain cooperative behavior. Mark, Noah. 1998. Beyond individual differences: Social differentiation from first principles. American Sociological Review 63: 309-330. This article uses an Agent-Based Model to show how, in a context of social structure and communication, differentiation can emerge spontaneously in undifferentiated societies. Scott, Nick, Michael Livingston, Aaron Hart, James Wilson, David Moore, and Paul Dietze. 2016. SimDrink: An agent-based Netlogo model of young, heavy drinkers for conducting alcohol policy experiments. Journal of Artificial Societies and Social Simulation 19: [http://jasss.soc.surrey.ac.uk/19/1/10.html] This article shows how Agent-Based Models might offer credible support to policy design. Analyzing Agent-Based Models There are a number of different aspects to analyzing Agent-Based Models. The first is validation, assessing the match between real data and its simulated equivalent. The quality of match needed depends on context. For an Agent-Based Model in a novel area even qualitative similarity may be suggestive (see for example time series comparisons in Ahrweiler, et al. 2015). Ideally, in progressive research, new models would outperform the match of existing ones but the dominance of nonempirical research means that there don’t seem to be any examples of this process. The nearest (far from satisfactory) example is Chattoe-Brown 2014 which shows that a popular Agent-Based Model of opinion dynamics produces outputs looking nothing like corresponding real data but that it is at least possible to design an Agent-Based Model with qualitative similarity to that data. Another new opportunity offered by Agent-Based Modeling that is rarely found in existing sociology is the capacity for replication. Using published research to re-implement an existing Agent-Based Model not only makes it unlikely that results arise from bugs (programs not behaving as intended) but also permits a really high level of critical engagement with model properties and their implications: The debate between Will 2009 and Macy and Sato 2010 (and preceding references) provides an example. The robustness of Agent-Based Models can also be explored by sensitivity analysis, which involves varying parameter values to see what effect this has on system behavior (Ten Broeke, et al. 2016). Ideally, calibration effort should be focused on parameters that most affect this behavior. Under certain circumstances sociological investigation may require an Agent-Based Model representing a rich environment within which agents have to operate (in rural sociology for example). Moreira et al. (2009) provide a discussion of so-called model linking where the Agent-Based Model [oxbibsubmitdefin.docx] 9 and another simulation (for example of climate) are of very different kinds and it makes sense to interface them with each other. By contrast Heckbert (2013) illustrates how environmental aspects (admittedly simplified) can be incorporated directly into an Agent-Based Model. Finally, there is best practice that can improve the progressive nature of Agent-Based Modelling. For example, organization of parameter information into a single table (see Scott, et al. 2016) makes it much clearer to what extent an Agent-Based Model has been effectively calibrated and focuses attention on further improvements: Replacing arbitrary values with those based on evidence and wide parameter ranges with narrower ones. Ahrweiler, Petra, Michel Schilperoord, Andreas Pyka, and Nigel Gilbert. 2015. Modelling research policy: Ex-ante evaluation of complex policy instruments. Journal of Artificial Societies and Social Simulation 18: [http://jasss.soc.surrey.ac.uk/18/4/5.html] An interesting research area with an unusual empirical approach. The model is calibrated on data from EU funding programs FP1-FP6 and then validated on FP7 data. Chattoe-Brown, Edmund. 2014. Using agent based modelling to integrate data on attitude change. Sociological Research Online 19: [http://www.socresonline.org.uk/19/1/16.html] Confronts a popular model of opinion dynamics with relevant data and then tries to develop an alternative Agent-Based Model, integrating insights from different disciplines, that displays a qualitative match with the same data. Heckbert, Scott. 2013. MayaSim: An agent-based model of the ancient Maya social-ecological system. Journal of Artificial Societies and Social Simulation 16: http://jasss.soc.surrey.ac.uk/16/4/11.html] This article implements an Agent-Based Model where agents have to operate socially in a geographically rich environment. Macy, Michael and Yoshimichi Sato. 2010. The surprising success of a replication that failed. Journal of Artificial Societies and Social Simulation 13: [http://jasss.soc.surrey.ac.uk/13/2/9.html] In conjunction with Will, this article illustrates the kind of detailed analysis and critique that can arise from replication. Moreira, Evaldinolia, Sérgio Costa, Ana Paula Aguiar, Gilberto Câmara, and Tiago Carneiro. 2009. Dynamical coupling of multiscale land change models. Landscape Ecology, 24: 1183-1194. [doi:10.1007/s10980-009-9397-x] This article provides an example that links different kinds of simulations (including AgentBased Modeling) to understand social behavior in a geographically complex environment. [oxbibsubmitdefin.docx] 10 Ten Broeke, Guus, George van Voorn, and Arend Ligtenberg. 2016. Which sensitivity analysis method should I use for my agent-based model? Journal of Artificial Societies and Social Simulation 19: [http://jasss.soc.surrey.ac.uk/19/1/5.html] This article provides a useful review and methodological discussion of sensitivity analysis for Agent-Based Modeling. Scott, Nick, Michael Livingston, Aaron Hart, James Wilson, David Moore, and Paul Dietze. 2016, SimDrink: An agent-based NetLogo model of young, heavy drinkers for conducting alcohol policy experiments. Journal of Artificial Societies and Social Simulation 19: [http://jasss.soc.surrey.ac.uk/19/1/10.html] Appendix A shows a calibration table, a newly emerging best practice to ensure that the calibration status of an Agent-Based Model (and how it can be further improved) is immediately clear. Will, Oliver. 2009. Resolving a replication that failed: News on the Macy and Sato model. Journal of Artificial Societies and Social Simulation 12: [http://jasss.soc.surrey.ac.uk/12/4/11.html] In conjunction with Macy and Sato, this article illustrates the kind of detailed analysis and critique that can be based on replication. Cognitive and Environmental Elements of Agent-Based Models This section introduces reviews and examples illustrating the range of key cognitive and environmental elements that can be found in most Agent-Based Models (though it is rare for any example to feature all of them.) Any Agent-Based Model has to specify how agents make decisions and, supporting that, how they perceive, categorize and recall information. Balke and Gilbert 2014 provide a recent survey that serves as a useful starting point. One advantage of Agent-Based Modeling is its ability to represent significantly different assumptions about the decision making process, for example neural networks (Beltratti et al. 1996) or evolutionary algorithms (Edmonds 1999). Other relevant aspects of cognition are less often represented in Agent-Based Models. For example, normative behavior is empirically important but modeled relatively rarely. Elsenbroich and Gilbert 2014 provide a thorough introduction. Another example of a relatively neglected area is complex communication, in which agents do more than simply report their current state. Dykstra, et al. 2013 provides an example of an Agent-Based Model in which agents argue and assess the communicative outcomes. Another key area in any Agent-Based Model is the representation of geographical space. Some simple examples still have agents wandering on a featureless plain but more sophisticated representations are often sociologically appropriate. Kim, et al. 2016 uses a geographically explicit Agent-Based Model to explore the interplay between geographical, social and epidemiological issues in the transmission of disease. Patel, et al. 2012 represent city growth and change to investigate the spatial distribution of slum formation. Pluchino, et al. 2014 offer a distinctive Agent-Based Model with [oxbibsubmitdefin.docx] 11 an explicit representation of indoor (rather than outdoor) space, representing visitor movement round a museum. Such examples are relevant not only for practical purposes (like emergency evacuation) but are also theoretically relevant to social mixing in organizations that has implications for innovation and control. Balke, Tina and Nigel Gilbert. 2014. How do agents make decisions? A survey. Journal of Artificial Societies and Social Simulation 17: [http://jasss.soc.surrey.ac.uk/17/4/13.html] This is a very thorough investigation of different decision-making architectures also providing extensive references. Beltratti, Andrea, Sergio Margarita, and Pietro Terna. 1996. Neural Networks for Economic and Financial Modelling. London: ITCP. Provides diverse examples of agents learning and interacting on the basis of neural network architectures that represent sophisticated learning (as opposed to rationality or rule based behavior). Dykstra, Piter, Corinna Elsenbroich, Wander Jager, Gerard Renardel de Lavalette, and Rineke Verbrugge. 2013. Put your money where your mouth is: DIAL, A dialogical model for opinion dynamics. Journal of Artificial Societies and Social Simulation 16: [http://jasss.soc.surrey.ac.uk/16/3/4.html] This is a rare example of an Agent-Based Model where communication involves more than just reporting agent internal states like opinions. Edmonds, Bruce. 1999. Gossip, sexual recombination and the El Farol Bar: Modelling the emergence of heterogeneity. Journal of Artificial Societies and Social Simulation 2: [http://jasss.soc.surrey.ac.uk/2/3/2.html] Shows how communication and heterogeneity can arise based on an evolutionary account. The El Farol Bar is a famous example of a stylized case where homogenous behavior is individually counter productive. Elsenbroich, Corinna and Nigel Gilbert. 2014. Modelling Norms. Berlin: Springer. Provides both a thorough overview of existing research and sample Agent-Based Models in this relatively neglected area of cognition. Kim, Hyeyoung, Ningchuan Xiao, Mark Moritz, Rebecca Garabed, and Laura Pomeroy. 2016. Simulating the transmission of foot-and-mouth disease among mobile herds in the Far North Region, Cameroon. Journal of Artificial Societies and Social Simulation 19: [http://jasss.soc.surrey.ac.uk/19/2/6.html] To understand disease transmission, Agent-Based Modeling is used to integrate knowledge of geography, social practices (in this case nomadic pastoralism) and epidemiology. [oxbibsubmitdefin.docx] 12 Patel, Amit, Andrew Crooks, and Naoru Koizumi. 2012. Slumulation: An agent-based modeling approach to slum formations. Journal of Artificial Societies and Social Simulation 15: [http://jasss.soc.surrey.ac.uk/15/4/2.html] This article illustrates an explicit representation of a complex geographical space (involving structuration by existing buildings) to explore the evolution of slums. Pluchino, Alessandro, Cesare Garofalo, Giuseppe Inturri, Andrea Rapisarda, and Matteo Ignaccolo. 2014. Agent-based simulation of pedestrian behaviour in closed spaces: A museum case study. Journal of Artificial Societies and Social Simulation 17: [http://jasss.soc.surrey.ac.uk/17/1/16.html] Geographically explicit representations need not be limited to models of outdoor activities. This article provides an indicative example for an indoor environment. Interactional and Social Elements of Agent-Based Models This section introduces reviews and examples illustrating the range of key interactional and social elements that can be found in most Agent-Based Models. These elements can be divided into three areas. The first is representations of organizations for which Chang and Harrington (2006) provide a thorough overview. There is also a journal (Computational and Mathematical Organization Theory) specializing in this area. However, the technique can also be used to represent other kinds of organizations. For example, Smaldino, et al. 2012 study informal groups and Chattoe 2006 analyzes voluntary membership organizations. The second key area involves interaction processes. Hammond and Axelrod (2006) model the evolution of ethnocentric behavior (differential treatment of one’s own kind) through repeated interaction while Sutcliffe and Wang (2012) looks at the mutual influences of trust and interaction on formation of social ties. This article could be viewed as complementary to Hummon 2000 that, while somewhat dated technically, is still an interesting and thorough investigation of the interplay between individual actions and aggregate social networks. Agent-Based Modeling can thus contribute to our understanding of both structured and unstructured social interactions. The final area involves different kinds of population change associated with sociological analysis of demography and migration. Hills and Todd 2008 model the process of marriage formation (and dissolution) while Kniveton, et al. 2011 provides a relatively rare published example of Agent-Based Modeling applied to migration processes. Chang, Myong-Hun and Joseph E. Harrington, Jr. 2006. “Agent-based models of organizations.” In Handbook of Computational Economics II: Agent-Based Computational Economics. Edited by Leigh Tesfatsion and Kenneth Judd, 1273-1337. Amsterdam: North-Holland. This chapter offers a thorough and relatively recent review of organizational models. [oxbibsubmitdefin.docx] 13 Chattoe, Edmund. 2006. Using simulation to develop testable functionalist explanations: A case study of church The survival. British Journal of Sociology 57: 379-397. [doi:10.1111/j.1468- 4446.2006.00116.x] This article analyzes churches as voluntary membership organizations (which must achieve balance between inflow and outflow of resources such as time and money). Hammond, Ross and Robert Axelrod. 2006. The evolution of ethnocentrism. Journal of Conflict Resolution 50: 926-936. This article illustrates Agent-Based Modeling to explore social evolution of interaction patterns in the long run. Hills, Thomas and Peter Todd. 2008. Population heterogeneity and individual differences in an assortative agent-based marriage and divorce model (MADAM) using search with relaxing expectations. Journal of Artificial Societies and Social Simulation 11: [http://jasss.soc.surrey.ac.uk/11/4/5.html] This article provides a demographic example of the interplay between individual choice (in this case partner choice) and aggregate outcomes. Hummon, Norman 2000. Utility and dynamic social networks. Social Networks 22: 221-249. [doi:10.1016/S0378-8733(00)00024-1] An early but insightful example of Agent-Based Modeling applied to social network evolution. Kniveton, Dominic, Christopher Smith, and Sharon Wood. 2011. Agent-based model simulations of future changes in migration flows for Burkina Faso. Global Environmental Change 21: S34-S40. [doi:10.1016/j.gloenvcha.2011.09.006] This article shows how decision-making and climate change scenarios can be integrated to project migration flows using an Agent-Based Model. Smaldino, Paul, Cynthia Pickett, Jeffrey Sherman, and Jeffrey Schank. 2012. An agent-based model of social identity dynamics. Journal of Artificial Societies and Social Simulation 15: [http://jasss.soc.surrey.ac.uk/15/4/7.html] This article explores the emergence of social groups from individual preferences for group membership and interaction. Sutcliffe, Alistair and Di Wang. 2012. Computational modelling of trust and social relationships. Journal of Artificial Societies and Social Simulation 15: [http://jasss.soc.surrey.ac.uk/15/1/3.html] Uses an Agent-Based Model to explore the interplay between social interactions and trust relationships. [oxbibsubmitdefin.docx] 14 Data and Agent-Based Models In terms of data use, Agent-Based Modeling is an odd mixture of the familiar and unfamiliar. Quantitative methods are accustomed to comparing one set of data with another so whether the comparators are statistical models or simulated data makes little difference. By contrast, qualitative methods give relatively little attention to integrated theory building and Agent-Based Modeling has not yet established a consensus on systematic use of qualitative data (but see Dilaver 2015, Ghorbani, et al. 2015). The ability of Agent-Based Modeling to represent complex and interdisciplinary social processes (and an aspiration to policy relevance) has also given rise to the distinct approach of companion/participatory modeling (Barnaud, et al. 2013). In this approach there is a balance to be struck between an Agent-Based Model that is empirically supported and one that has the endorsement of relevant stakeholders in the social process. Another area where data is particularly relevant is in the establishment of progressive research practice. In a useful debate, advocates of Agent-Based Modeling (Fossett 2011) are critiqued for lack of engagement with existing data (Goering 2006) while different empirically inspired approaches, despite their individual merits, are not well integrated into a progressive research program (Bruch and Mare 2006, Hatna and Benenson 2012). Finally, there is an interesting relationship between types of data and the available processes of calibration and validation. Although calibration can be done using traditional quantitative methods (for validation against aggregate statistics because these are the kinds of data sociology typically has available), it may be possible to independently calibrate and validate Agent-Based Models at the individual and small group level using laboratory experiments (Arthur 1993). Arthur, Brian. 1993. On designing economic agents that behave like human agents. Journal of Evolutionary Economics 3: 1-22. Although this does not involve a sociological application, the general point that calibration and validation are not limited to whole social systems is important. Barnaud, Cécile, Christophe Le Page, Pongchai Dumrongrojwatthana, and Guy Trébuil. 2013. Spatial representations are not neutral: Lessons from a participatory agent-based modelling process in a landuse conflict. Environmental Modelling and Software 45: 150-159. [doi: 10.1016/j.envsoft.2011.11.016] A novel illustration of the participatory/companion modeling approach to the integration of data collection and model building. Bruch, Elizabeth and Robert Mare. 2006. Neighborhood choice and neighborhood change. American Journal of Sociology, 112: 667-709. Apart from a specific contribution to progressive research on ethnic residential segregation, this article also shows how data collection can be tailored to Agent-Based Modeling approaches. [oxbibsubmitdefin.docx] 15 Dilaver, Ozge. 2015. From participants to agents: Grounded simulation as a mixed-method research Journal design. of Artificial Societies and Social Simulation 18: [http://jasss.soc.surrey.ac.uk/18/1/15.html] This is one possible attempt to systematize the use of qualitative data in constructing AgentBased Models where no consensus yet exists. Fossett, Mark. 2011. Generative models of segregation: Investigating model-generated patterns of residential segregation by ethnicity and socioeconomic status. The Journal of Mathematical Sociology 35: 114-145. [doi:10.1080/0022250X.2010.532367] This article attempts to extend the Schelling approach but its commitment to calibration and validation could be questioned. Hatna, Erez and Itzhak Benenson. 2012. The Schelling Model of ethnic residential dynamics: Beyond the integrated-segregated dichotomy of patterns. Journal of Artificial Societies and Social Simulation, 15: [http://jasss.soc.surrey.ac.uk/15/1/6.html] This article presents a different development of the Schelling approach involving the use of real residential data. Ghorbani, Amineh, Gerard Dijkema, and Noortje Schrauwen. 2015. Structuring qualitative data for agent-based modelling. Journal of Artificial Societies and Social Simulation 18: [http://jasss.soc.surrey.ac.uk/18/1/2.html] This is a different attempt to systematize the use of qualitative data in constructing AgentBased Models where no consensus yet exists. Goering, John. 2006. ‘Shelling [sic] redux: How sociology fails to make progress in building and empirically testing complex causal models regarding race and residence. The Journal of Mathematical Sociology 30: 299-317. [doi:10.1080/00222500500544144] This article is a detailed critique of Fossett (and effectively of Agent-Based Modeling practice more generally). It is useful for the reader to access debates of this kind that surface arguments for and against the approach that sometimes remain implicit. Institutions of Agent-Based Modeling Six different institutional dimensions of Agent-Based Modeling can be identified that are relevant to this bibliography. The first concerns the relationship between the approach and the institution of sociology itself. The ability to represent complex systems and to avoid assumptions made for technical reasons alone allows Agent-Based Models to be less anchored in specific disciplines. Given its subject matter (linking individual criminal behavior to observed patterns of crime statistics), it is not clear how useful it is to try and label Bosse and Gerritsen 2010 as being (or not being) sociology, but it is definitely a [oxbibsubmitdefin.docx] 16 valuable article. In the same manner, the debate between Aron 1988 and Kalick and Hamilton 1988 (and prior references) on a model of mate choice provides an excellent illustration of the Agent-Based Modeling methodology in action but while attractiveness might be considered a psychological concept, marriage has a clear sociological relevance (and it would be odd to claim that attractiveness didn’t matter in the face of the evidence just because it happened to be part of psychology). More generally (Billari and Prskawetz 2013), Agent-Based Modeling has now matured to the point where more specialized application areas are being established. The second, related to the first, is the extent to which Agent-Based Modeling can be differentially associated with certain areas or approaches to sociology. One notable example is Analytical Sociology (Manzo 2014a), which has adopted the approach as a key element of its methodological palette. The third, to be mentioned only briefly, is software choice. There are numerous programming languages that support Agent-Based Modeling, varying widely in technical accessibility and the size of their user communities. Developments in this area are rapid and general advice that does not take account of the skills and selection criteria of the user dates very rapidly. Kravari and Bassiliades 2015 provide a serviceable recent overview. The fourth, also rather brief, is the existence of online resources (searchable via google). A typical example is OpenABM, which provides features like a models library, lists of relevant journals, some teaching resources and a bibliography. However, like many community web sites, it is hard to predict its long-term robustness and coverage is idiosyncratic. There does not seem to be a single really high quality web resource for ABM. The fifth is the publication of Agent-Based Modeling in languages other than English. As with many other academic fields, non-English publications are rather scarce and themselves frequently cite a great majority of English sources on the topic (suggesting that the issue is a genuine lack of relevant sources rather than Anglophone bias). Nonetheless, cosmopolitan readers may wish to follow up indicative examples of relevant work in French (Manzo 2014b), German (Flache and Mäs 2015) and Italian (Gabriellini 2011) that provide (slightly) better access to non English sources. Finally, as a maturing approach, Agent-Based Modeling is starting to develop self-awareness of its wider impact and diffusion. Articles like Squazzoni and Casnici 2013 are useful in giving a sense of where, and to what extent, the approach is represented in sociological (and for that matter other) journals. Aron, Arthur. 1988. The matching hypothesis reconsidered again: Comment on Kalick and Hamilton. Journal of Personality and Social Psychology 54: 441-446. [doi:10.1037/0022-3514.54.3.441] [oxbibsubmitdefin.docx] 17 This is a further useful example of dialogue to establish (or reject) the effectiveness of AgentBased Modeling in a substantive domain. Billari, Francesco and Alexia Prskawetz, eds. 2013. Agent-based computational demography: Using simulation to improve our understanding of demographic behaviour Heidelberg: Physica. This book illustrates the recent development of more specialized sub fields in Agent-Based Modeling. Bosse, Tibor and Charlotte Gerritsen. 2010. Social simulation and analysis of the dynamics of criminal hot spots. Journal of Artificial Societies and Social Simulation 13: [http://jasss.soc.surrey.ac.uk/13/2/5.html] Another illustration of the methodology of Agent-Based Modeling applied to a different domain of social science. Flache, Andreas and Michael Mäs. 2015. “Multi-Agenten-Modelle.” In Handbuch Modellbildung und Simulation in den Sozialwissenschaften. Edited by Norman Braun and Nicole Saam, 491-514. Wiesbaden: Springer. A useful recent overview of Agent-Based Modeling published in German. Gabriellini, Simone. 2011. Simulare meccanismi sociali con NetLogo: Una introduzione. Milan: Franco Angeli. A recent overview combining guidance in Netlogo programming and sociologically relevant examples, published in Italian. Kalick, Michael and Thomas Hamilton. 1988. Closer look at a matching simulation: Reply to Aron. Journal of Personality and Social Psychology 54: 447-451. [doi:10.1037/0022-3514.54.3.447] This is another useful example of dialogue to establish (or reject) the effectiveness of AgentBased Modeling in a substantive domain. Kravari, Kalliopi and Nick Bassiliades. 2015. A survey of agent platforms. Journal of Artificial Societies and Social Simulation 18: [http://jasss.soc.surrey.ac.uk/18/1/11.html] Software packages rise and fall rapidly in this field. This article provides effective coverage at the time of writing but unfortunately will probably date relatively rapidly. Manzo, Gianluca, ed. 2014a. Analytical Sociology: Actions and Networks. Chichester: Wiley. The chapters by Rolfe, González-Bailón, et al. and Gabbriellini provide examples of AgentBased Modeling associated with Analytical Sociology. Manzo, Gianluca 2014b. Potentialités et limites de la simulation multi-agents: Une introduction. Revue Française de Sociologie 55: 653-688. [doi:10.3917/rfs.554.0653] [oxbibsubmitdefin.docx] 18 A recent critical overview of Agent-Based Modeling published in French. Squazzoni, Flaminio and Niccolò Casnici. 2013. Is social simulation a social science outstation? A bibliometric analysis of the impact of JASSS. Journal of Artificial Societies and Social Simulation 16: [http://jasss.soc.surrey.ac.uk/16/1/10.html] This clearly emphasizes the journal in which it appears but nonetheless gives a useful overview of the diffusion of Agent-Based Modeling because JASSS publishes so much more in this area than any other journal. Challenges to Agent-Based Modeling Challenges to Agent-Based Modeling fall into two broad groups that have a logical connection. Although the methodology of the approach naturally indicates how progressive research might work (new models should fit the old data better or fit more data or fit no worse but with reduced calibration uncertainty), the minority status of calibrated and validated models has put this logic into practical abeyance. Thus we observe how research in Agent-Based Modeling typically develops by permuting various elements of models such as networks (Chiang 2013), adaptation (Izquierdo, et al. 2008) and geography (Power 2009) without putting these variants to empirical test. (All of these examples involve models of interaction in repeated games.) This problem might be referred to as validation ambiguity. This leads to the second challenge. While it appears intuitive that certain neglected aspects of AgentBased Models are sociologically relevant, the danger is that such intuitions remain just intuitions (with merely not being totally implausible as a very weak filtering criterion) or lead to a yet larger space of non-progressive model variations. Relevant examples of such relatively neglected areas include sophisticated communication (but see Mastrangeli, et al. 2010), context – the fact that social actors may behave differently in different company or locations (but see Edmonds 2001), open ended social change – which involves the arrival of new individuals, technologies, social practices and so on during a simulation run (but see Watts and Gilbert 2014) and reputation (Conte and Paolucci 2002). Another example, linked to the previous discussion concerning Agent-Based Models of repeated games would be the idea of choice and refusal of partners (Stanley, et al. 1994), which seems far more rare in published models than its apparent sociological relevance would suggest. This problem might be referred to as calibration ambiguity. This bibliography thus ends as it began, with an emphasis on the importance of effective methodology in ensuring good quality research. Chiang, Yen-Sheng. 2013. Cooperation could evolve in complex networks when activated conditionally on network characteristics. Journal of Artificial Societies and Social Simulation 16: [http://jasss.soc.surrey.ac.uk/16/2/6.html] This article emphasizes the network element in explaining cooperation in repeated games. [oxbibsubmitdefin.docx] 19 Conte, Rosaria and Mario Paolucci. 2002. Reputation in Artificial Societies: Social Beliefs for Social Order. Dordrecht: Kluwer Academic Publishers. This book provides an overview of theories and models of reputation, a phenomenon that seems more sociologically important than its appearance in Agent-Based Models suggests. Edmonds, Bruce. 2001. “Learning appropriate contexts.” In Modeling and using context. Edited by Varol Akman, Paolo Bouquet, Richmond Thomason, and Roger Young, 143-155. Berlin: Springer. This chapter provides a useful discussion of context (different behavior in different settings), which seems more sociologically important than the current emphasis in Agent-Based Modeling would suggest. Izquierdo, Segismundo, Luis Izquierdo, and Nicholas Gotts. 2008. Reinforcement learning dynamics in social dilemmas. Journal of Artificial Societies and Social Simulation 11: [http://jasss.soc.surrey.ac.uk/11/2/1.html] This article emphasizes the adaptive element in explaining cooperation in repeated games. Mastrangeli, Massimo, Martin Schmidt, and Lucas Lacasa. 2010. The roundtable: An abstract model of conversation dynamics. Journal of Artificial Societies and Social Simulation 13: [http://jasss.soc.surrey.ac.uk/13/4/2.html] This article uses Agent-Based Modeling to reflect more sophisticated features of conversation like turn taking. Power, Conrad. 2009. A spatial agent-based model of n-person Prisoner’s Dilemma cooperation in a socio-geographic community. Journal of Artificial Societies and Social Simulation 12: [http://jasss.soc.surrey.ac.uk/12/1/8.html] This article emphasizes the spatial element in explaining cooperation in repeated games. Stanley, Ann, Dan Ashlock, and Leigh Tesfatsion 1994. “Iterated Prisoner’s Dilemma with choice and refusal of partners.” In Artificial Life III. Edited by Christopher G. Langton, 131-175. Reading, MA: Addison-Wesley. This chapter illustrates the phenomenon of choice and refusal (who you play with rather than how you play), which seems more sociologically relevant than its appearance in Agent-Based Models suggests. Watts, Christopher and Nigel Gilbert. 2014. Simulating innovation: Computer-based tools for rethinking innovation. Cheltenham, UK: Edward Elgar. This book provides a thorough overview of Agent-Based Models connected with innovation with chapter 7 focusing on a distinctive model representing open-ended change.