Joshua M. Epstein

Agent-based computational models and generative social science

This article argues that the agent-based computational model permits a distinctive approach to social science for which the term “generative” is suitable. In defending this terminology, features distinguishing the approach from both “inductive” and “deductive” science are given. Then, the following specific contributions to social science are discussed: The agent-based computational model is a new tool for empirical research. It offers a natural environment for the study of connectionist phenomena in social science. Agent-based modeling provides a powerful way to address certain enduring—and especially interdisciplinary—questions. It allows one to subject certain core theories—such as neoclassical microeconomics—to important types of stress (e.g., the effect of evolving preferences). It permits one to study how rules of individual behavior give rise—or “map up”—to macroscopic regularities and organizations. In turn, one can employ laboratory behavioral research findings to select among competing agent-based (“bottom up”) models. The agent-based approach may well have the important effect of decoupling individual rationality from macroscopic equilibrium and of separating decision science from social science more generally. Agent-based modeling offers powerful new forms of hybrid theoretical-computational work; these are particularly relevant to the study of non-equilibrium systems. The agentbased approach invites the interpretation of society as a distributed computational device, and in turn the interpretation of social dynamics as a type of computation. This interpretation raises important foundational issues in social science—some related to intractability, and some to undecidability proper. Finally, since “emergence” figures prominently in this literature, I take up the connection between agent-based modeling and classical emergentism, criticizing the latter and arguing that the two are incompatible. ! 1999 John Wiley & Sons, Inc.

Multiple myeloma disrupts the TRANCE/ osteoprotegerin cytokine axis to trigger bone destruction and promote tumor progression

Bone destruction, caused by aberrant production and activation of osteoclasts, is a prominent feature of multiple myeloma. We demonstrate that myeloma stimulates osteoclastogenesis by triggering a coordinated increase in the tumor necrosis factor-related activation-induced cytokine (TRANCE) and decrease in its decoy receptor, osteoprotegerin (OPG). Immunohistochemistry and in situ hybridization studies of bone marrow specimens indicate that in vivo, deregulation of the TRANCE–OPG cytokine axis occurs in myeloma, but not in the limited plasma cell disorder monoclonal gammopathy of unknown significance or in nonmyeloma hematologic malignancies. In coculture, myeloma cell lines stimulate expression of TRANCE and inhibit expression of OPG by stromal cells. Osteoclastogenesis, the functional consequence of increased TRANCE expression, is counteracted by addition of a recombinant TRANCE inhibitor, RANK-Fc, to marrow/myeloma cocultures. Myeloma–stroma interaction also has been postulated to support progression of the malignant clone. In the SCID-hu murine model of human myeloma, administration of RANK-Fc both prevents myeloma-induced bone destruction and interferes with myeloma progression. Our data identify TRANCE and OPG as key cytokines whose deregulation promotes bone destruction and supports myeloma growth.

Aligning simulation models: A case study and results

This paper develops the concepts and methods of a process we will call “alignment of computational models” or “docking” for short. Alignment is needed to determine whether two models can produce the same results, which in turn is the basis for critical experiments and for tests of whether one model can subsume another. We illustrate our concepts and methods using as a target a model of cultural transmission built by Axelrod. For comparison we use the Sugarscape model developed by Epstein and Axtell.The two models differ in many ways and, to date, have been employed with quite different aims. The Axelrod model has been used principally for intensive experimentation with parameter variation, and includes only one mechanism. In contrast, the Sugarscape model has been used primarily to generate rich “artificial histories”, scenarios that display stylized facts of interest, such as cultural differentiation driven by many different mechansims including resource availability, migration, trade, and combat.The Sugarscape model was modified so as to reproduce the results of the Axelrod cultural model. Among the questions we address are: what does it mean for two models to be equivalent, how can different standards of equivalence be statistically evaluated, and how do subtle differences in model design affect the results? After attaining a “docking” of the two models, the richer set of mechanisms of the Sugarscape model is used to provide two experiments in sensitivity analysis for the cultural rule of Axelrods model.Our generally positive experience in this enterprise has suggested that it could be beneficial if alignment and equivalence testing were more widely practiced among computational modelers.

Controlling Pandemic Flu: The Value of International Air Travel Restrictions

Background Planning for a possible influenza pandemic is an extremely high priority, as social and economic effects of an unmitigated pandemic would be devastating. Mathematical models can be used to explore different scenarios and provide insight into potential costs, benefits, and effectiveness of prevention and control strategies under consideration. Methods and Findings A stochastic, equation-based epidemic model is used to study global transmission of pandemic flu, including the effects of travel restrictions and vaccination. Economic costs of intervention are also considered. The distribution of First Passage Times (FPT) to the United States and the numbers of infected persons in metropolitan areas worldwide are studied assuming various times and locations of the initial outbreak. International air travel restrictions alone provide a small delay in FPT to the U.S. When other containment measures are applied at the source in conjunction with travel restrictions, delays could be much longer. If in addition, control measures are instituted worldwide, there is a significant reduction in cases worldwide and specifically in the U.S. However, if travel restrictions are not combined with other measures, local epidemic severity may increase, because restriction-induced delays can push local outbreaks into high epidemic season. The per annum cost to the U.S. economy of international and major domestic air passenger travel restrictions is minimal: on the order of 0.8% of Gross National Product. Conclusions International air travel restrictions may provide a small but important delay in the spread of a pandemic, especially if other disease control measures are implemented during the afforded time. However, if other measures are not instituted, delays may worsen regional epidemics by pushing the outbreak into high epidemic season. This important interaction between policy and seasonality is only evident with a global-scale model. Since the benefit of travel restrictions can be substantial while their costs are minimal, dismissal of travel restrictions as an aid in dealing with a global pandemic seems premature.

Learning to Be Thoughtless: Social Norms and Individual Computation

This paper extends the literature on the evolution of norms with anagent-based modelcapturing a phenomenon that has been essentially ignored, namely thatindividual thought – orcomputing – is often inversely related to the strength of a social norm.Once a norm isentrenched, we conform thoughtlessly. In this model, agents learn how tobehave (what normto adopt), but – under a strategy I term Best Reply to Adaptive SampleEvidence – they also learnhow much to think about how to behave. How much they are thinking affects howthey behave,which – given how others behave – affects how much they think. Inshort, there is feedbackbetween the social (inter-agent) and internal (intra-agent) dynamics. Inaddition, we generate thestylized facts regarding the spatio-temporal evolution of norms: localconformity, global diversity,and punctuated equilibria.

Modelling to contain pandemics.

Agent-based computational models can capture irrational behaviour, complex social networks and global scale--all essential in confronting H1N1, says Joshua M. Epstein.

Coupled Contagion Dynamics of Fear and Disease: Mathematical and Computational Explorations

Background In classical mathematical epidemiology, individuals do not adapt their contact behavior during epidemics. They do not endogenously engage, for example, in social distancing based on fear. Yet, adaptive behavior is well-documented in true epidemics. We explore the effect of including such behavior in models of epidemic dynamics. Methodology/Principal Findings Using both nonlinear dynamical systems and agent-based computation, we model two interacting contagion processes: one of disease and one of fear of the disease. Individuals can “contract” fear through contact with individuals who are infected with the disease (the sick), infected with fear only (the scared), and infected with both fear and disease (the sick and scared). Scared individuals–whether sick or not–may remove themselves from circulation with some probability, which affects the contact dynamic, and thus the disease epidemic proper. If we allow individuals to recover from fear and return to circulation, the coupled dynamics become quite rich, and can include multiple waves of infection. We also study flight as a behavioral response. Conclusions/Significance In a spatially extended setting, even relatively small levels of fear-inspired flight can have a dramatic impact on spatio-temporal epidemic dynamics. Self-isolation and spatial flight are only two of many possible actions that fear-infected individuals may take. Our main point is that behavioral adaptation of some sort must be considered.

REMARKS ON THE FOUNDATIONS OF AGENT-BASED GENERATIVE SOCIAL SCIENCE †

This chapter treats a variety of epistemological issues surrounding generative explanation in the social sciences, and discusses the role of agent-based computational models in generative social science.

A Distributed Platform for Global-Scale Agent-Based Models of Disease Transmission

The Global-Scale Agent Model (GSAM) is presented. The GSAM is a high-performance distributed platform for agent-based epidemic modeling capable of simulating a disease outbreak in a population of several billion agents. It is unprecedented in its scale, its speed, and its use of Java. Solutions to multiple challenges inherent in distributing massive agent-based models are presented. Communication, synchronization, and memory usage are among the topics covered in detail. The memory usage discussion is Java specific. However, the communication and synchronization discussions apply broadly. We provide benchmarks illustrating the GSAM’s speed and scalability.

A hybrid epidemic model: combining the advantages of agent-based and equation-based approaches

Agent-based models (ABMs) are powerful in describing structured epidemiological processes involving human behavior and local interaction. The joint behavior of the agents can be very complex and tracking the behavior requires a disciplined approach. At the same time, equation- based models (EBMs) can be more tractable and allow for at least partial analytical insight. However, inadequate representation of the detailed population structure can lead to spurious results, especially when the epidemic process is beginning and individual variation is critical. In this paper, we demonstrate an approach that combines the two modeling paradigms and introduces a hybrid model that starts as agent-based and switches to equation-based after the number of infected individuals is large enough to support a population-averaged approach. This hybrid model can dramatically save computational times and, more fundamentally, allows for the mathematical analysis of emerging structures generated by the ABM.