Search Results

This chapter provides an overview of ways to create agent-based models, including agent-based modeling and simulation architectures and implementation tools. It also discusses model growth paths for enhancing systems and related issues.

This chapter extends previous work analyzing the origins of academic entrepreneurship at Stanford with an agent-based model that simulates the rise and spread of patenting by research faculty, drawing on archival analysis of divergent approaches taken by different lab directors. In so doing, this chapter builds on the formal model of autocatalysis developed in Chapter 3, which enables this chapter to disentangle competing explanations. The results are quite surprising. Incentives or mimicry alone are less likely to account for academic embrace of patenting, whereas preemptive efforts to preserve scientific autonomy do play a large role. The pursuit of safeguards from commercial co-optation by other researchers has the transformative effect of making the emergence of proprietary science more likely.

This chapter begins with a brief discussion of the need for a new approach to modeling party competition. It then makes a case for the use of agent-based modeling to study multiparty competition in an evolving dynamic party system, given the analytical intractability of the decision-making environment, and the resulting need for real politicians to rely on informal decision rules. Agent-based models (ABMs) are “bottom-up” models that typically assume settings with a fairly large number of autonomous decision-making agents. Each agent uses some well-specified decision rule to choose actions, and there may be considerable diversity in the decision rules used by different agents. Given the analytical intractability of the decision-making environment, the decision rules that are specified and investigated in ABMs are typically based on adaptive learning rather than forward-looking strategic analysis, and agents are assumed to have bounded rather than perfect rationality. An overview of the subsequent chapters is also presented.

This concluding chapter summarizes key themes and presents some final thoughts. This book started with the twin premises that understanding multiparty competition is a core concern for everyone interested in representative democracy and that multiparty competition should be understood as an evolving dynamic system, not a stationary state. Given these premises, it investigated the dynamics of multiparty competition using computational agent-based modeling, a new technology that is ideally suited to providing systematic answers to the types of question we want to ask. This allows the modeling of decision making by party leaders, in what is clearly an analytically intractable setting, in terms of the informal rules of thumb that might be used by real human beings, rather than the formally provable best response strategies used by traditional formal theorists. Whether people use the dynamic model of multiparty competition or some better model of this vital but complex political process, there is no doubt that the computational approach deployed in this book offers vast potential to ask and answer interesting and important questions.

This chapter further develops an agent-based model of economic production from the previous chapter. It shows that certain limitations intrinsic to the original hypercycle model—in particular, complexity barriers and vulnerability to parasites—are overcome once autocatalysis takes place in a spatial context, rather than in random-topology liquids. Localized heterogeneity in spatial interaction induces the inscription of path dependencies into cells. This explains why life becomes enhanced once it is embodied. The model also demonstrates why altruism and stigmergy produce more complex rule-chemistries. Altruistic reproduction and stigmergy are superior to selfish reproduction and fixed environments, respectively, because of their superior capacities for self-repair. Beyond suggestive specifics, the hypercycle model and its extensions show how chemistry and economic production and trading in markets can be mapped onto each other, sparking insights for both sides.

This chapter develops the methods for designing, executing, and analyzing large suites of computer simulations that generate stable and replicable results. It starts with a discussion of the different methods of experimental design, such as grid sweeping and Monte Carlo parameterization. Next, it demonstrates how to calculate mean estimates of output variables of interest. It does so by first discussing stochastic processes, Markov Chain representations, and model burn-in. It focuses on three stochastic process representations: nonergodic deterministic processes that converge on a single state; nondeterministic stochastic processes for which a time average provides a representative estimate of the output variables; and nondeterministic stochastic processes for which a time average does not provide a representative estimate of the output variables. The estimation strategy employed depends on which stochastic process the simulation follows. Lastly, the chapter presents a set of diagnostic checks used to establish an appropriate sample size for the estimation of the means.

This chapter introduces agent-based modeling as a tool for studying the dynamics of ethnic demography. Computer simulations using agent-based models allow us to advance the static Boulet and Chandra framework. This chapter demonstrates how core concepts in the constructivist agenda such as “stickiness” of attributes and “salience” of dimensions can be modeled as continuous variables that are mutually constitutive of identity shifts. It first presents simulations that investigate the relationship between the stickiness of a salient attribute and the longevity of leadership of ethnic entrepreneurs who build a constituency based upon identification with that attribute. Next, it applies the model to examine the relationship of electoral thresholds to the concentration of attributes in both ethnic and non-ethnic demographies. The simulations reveal that the electoral system is a significant intervening factor in both of these applications. The chapter concludes by discussing the advantages of agent-based models for analyzing multi-layered and complex interactions.

The conceptual framework that this book gradually develops across the various chapters — I-Space — has been used on numerous occasions as conceptual aids both in research and in consulting interventions. It has proven its worth as a sensemaking device in complex situations. But does it have any predictive value? At present, it helps to understand rather than predict. To develop a capacity to predict, the framework would have to be further researched. This chapter therefore first briefly recapitulates the key points presented in the book's previous chapters, and then looks ahead at the kind of research agenda that these points imply. It then identifies several possible avenues of research, from mapping knowledge, cultural and institutional structures, and learning processes to further developing the agent-based simulation model that implements the main provisions of the theorizing underpinning the framework. The research agenda keeps growing. To engage with it, the I-Space Institute whose mission is to carry out I-Space-related research and consulting, has been founded.

This book is about the integration of agent-based modeling and network science. The leading chapter gives a brief historical review as a background and motivation of the book. The historical review begins with the network origin of agent-based models, the rising of autonomous agents and network sciences, the game-theoretic formulation of networks, and the agent-based formulation of networks, and ends with networks as an alternative manifestation of markets. Along this series of development, we see how network work becomes an indispensable element of agent-based models, and how agents-based models can be a powerful tool for modeling the dynamics on and of networks. With this large picture, the book is divided into three parts: foundations, primary dynamics on and of networks, and applications.

Agent-based modeling is a newer approach to the study of neighborhoods and health. In brief, an agent-based model is one of a class of computational models for simulating the actions and interactions of autonomous agents (both individual or collective entities, such as organizations or groups) with a view to assessing their effects on the system as a whole. Neighborhood characteristics and resources evolve and adapt as the individuals living within them change and vice versa. In this way, neighborhoods reflect a complex adaptive system. In this chapter, we introduce agent-based models as a tool for modeling these interactive and adaptive processes that occur within a system, such as a neighborhood. The chapter provides a basic introduction to this method, drawing on examples from the neighborhoods and health literature.

This chapter specifies the “baseline” agent-based model of dynamic multiparty competition, which derives from an article published by (Laver 2005). This assumes that each voter has in mind some personal ideal “package” of policy positions and supports the political party that offers the policy package closest to this. The dynamic system at the heart of the model is as follows: voters support their “closest” party in this sense; party leaders adapt the policy packages they offer in light of the revealed pattern of voter support; voters reconsider which party they support in light of the revealed pattern of party policy packages; and this process continues forever. This recursive model describes policy-based party competition as a complex system, and the baseline model specifies three decision rules that party leaders may deploy when they choose party policy positions in such a setting. These rules are Sticker (always keep the same position), Aggregator (move policy to the centroid of the ideal policy positions of your current supporters), and Hunter (if your last policy move increased your support, make another move in the same direction; or else change heading and move in a different direction).

As we have seen, numerous analysis and modeling tools that take into account the natural complexity of systems and dissemination and implementation processes are available, and the use of them is increasing over time. This chapter summarizes the characteristics, potential insights, and limitations of each modeling approach. It is important to note that modeling from a systems perspective, like all modeling approaches, requires assumptions about variables to include (or exclude), and hypothesized relationships dictate the quality of the model and the utility of the results. As such, using theory and empirical data to inform model design is paramount. Systems thinking and methods remain underutilized in dissemination and implementation despite demonstrations of the utility of incorporating systems thinking and methods into dissemination and implementation studies.

Chapter 2 reviews the development of the network-based agent-based models. From the behavioral and decision-making perspective of agents, the network-based agent-based model is accompanied by the neighborhood-based decision rules. The chapter divides the literature into two parts. The one developed before the advent of modern network science normally relies on the one-dimensional or two-dimensional lattices (cellular automata). The one developed with the advent of modern network science relies on the newly proposed network generation algorithms. In a chronological order, the chapter demonstrates the two-generation network-based agent-based models via a number of pioneering works. The purpose of these demonstrations is to show how network topologies can affect the operation of various economic and social systems, including residential segregation, pro-social behavior, oligopolistic competition, market sentiment, sharing of public resources, market mechanism, marketing, and macroeconomic stability. Cellular automata as the theoretical underpinning of undecidability and unpredictability for the dynamics on networks are also introduced.

Agent-based modeling (ABM) provides a viable approach for investigating complex phenomena. ABM computationally simulates the interactions of autonomous agents in order to assess their effect on whole systems and aids in visualizing that clients self-organize at the grassroots level. ABMs create a "social reality" generated from several inputs. This chapter provides the reader with essential background on developing an agent-based model, and instructions for downloading and installing the freely distributed NetLogo software. This supplementary material includes using the tab features to write, debug, and run code. Although more advanced software is available, NetLogo has the advantage of being user-friendly. Computational models are computer programs that we develop.

This chapter conveys the key concepts, overarching methods, and common applications of agent-based modeling in population health science. This chapter will also provide the reader with a foundational understanding of how and why agent-based models (ABMs) are increasingly employed to address pressing public health challenges of the 21st century. It demonstrates these concepts with an example that adopts ABM to simulate HIV transmission dynamics in high-risk populations. The chapter concludes with an in-depth discussion of two of the most promising avenues for the continued adoption of agent-based modeling approaches to improve population health: the evaluation of policy experiments and evidence synthesis.

This chapter begins the investigation of multiparty competition using the baseline model specified in Chapter 3 and methods and procedures specified in Chapter 4. The most significant results concern the representativeness of evolved configurations of party policy positions. In symmetric populations, the ideal points of voters are not best represented by a set of (Hunter) parties who compete for their support by trying to find popular policy positions. Instead, voter preferences are better represented by a set of (Aggregator) parties that do not compete with each other on policy at all but instead seek to represent the policy preferences only of their current supporters. This happens because the dynamics of vote-seeking competition in this setting cause parties to set policy positions closer to the center of the policy space than would be needed for optimal representation—while at the same time avoiding the dead center of the space.