Skip to main content

Über dieses Buch

This volume is a post-conference publication of the 4th World Congress on Social Simulation (WCSS), with contents selected from among the 80 papers originally presented at the conference. WCSS is a biennial event, jointly organized by three scientific communities in computational social science, namely, the Pacific-Asian Association for Agent-Based Approach in Social Systems Sciences (PAAA), the European Social Simulation Association (ESSA), and the Computational Social Science Society of the Americas (CSSSA). It is, therefore, currently the most prominent conference in the area of agent-based social simulation. The papers selected for this volume give a holistic view of the current development of social simulation, indicating the directions for future research and creating an important archival document and milestone in the history of computational social science. Specifically, the papers included here cover substantial progress in artificial financial markets, macroeconomic forecasting, supply chain management, bank networks, social networks, urban planning, social norms and group formation, cross-cultural studies, political party competition, voting behavior, computational demography, computational anthropology, evolution of languages, public health and epidemics, AIDS, security and terrorism, methodological and epistemological issues, empirical-based agent-based modeling, modeling of experimental social science, gaming simulation, cognitive agents, and participatory simulation. Furthermore, pioneering studies in some new research areas, such as the theoretical foundations of social simulation and categorical social science, also are included in the volume.



Online Communities and Social Media


Chapter 1. Stock BBS Factor Model Using Principal Component Score

Our aim is to develop a new factor for stock bulletin board system (BBS) postings that is different from our bullish-bearish model (BMB) factor. In our previous study, the content of stock BBS postings was classified into two categories, i.e., bullish postings and bearish postings, and our BMB factor is based on these categories. The results of recent studies suggest that the content of stock BBS postings may be represented by employing more than one index. To develop new factors on the basis of the principal component score, we use morphological analysis and principal component analysis to analyze the content of stock BBS postings. As a result, we find candidates for new factors that can explain stock returns.
Hirohiko Suwa, Eiichi Umehara, Toshizumi Ohta

Chapter 2. How Consumer-Generated Advertising Works: An Empirical Agent-Based Simulation

Affiliate advertising is a novel form of Internet advertising that enables bloggers to insert advertising for any product in their blog articles and to gain rewards based on consumers’ actual responses. To understand how this form of advertising works, we conducted Web-based questionnaire surveys among bloggers, including affiliates, and readers, including buyers. Moreover, we constructed an agent-based model that is empirically validated by the above data. The results of the simulation using this model showed that (1) link structures between affiliates and readers have a remarkable impact on the average revenues of all affiliates, and (2) the presence of random walkers in searching for a better ad mix may increase total revenues for all affiliates compared to the presence of local imitators. Based on these results, we discuss the managerial implications and suggest future avenues for research.
Makoto Mizuno

Chapter 3. Understanding Citizens’ Channel Choice of Public Service Delivery: An Agent-Based Simulation Approach

E-government has become a rapidly emerging research field in recent decades. An obvious benefit of deploying e-government service is that the quality is improved through integrated services and the services provided can more flexibly satisfy the citizens’ needs. However, the adoption rate of e-government is still relatively low, especially for transactional services. We propose a new method, agent-based simulation, to understand how citizens from different social groups choose channels to utilize certain kinds of governmental services over time. In addition, by evaluating different public strategies such as public propaganda and increased technical support, we try to identify what kind of strategy could attract more citizens to utilize e-government services.
Shuang Chang, Manabu Ichikawa, Hiroshi Deguchi

Chapter 4. Cyclical Pattern of the Rise and Fall of an Online Community Due to a Troll

A deviant behavior against an online community such as a troll sometimes forces a community to close. A troll’s action is regarded as a crime of pleasure: a selfish, anonymous individual feels delight in inflaming others and in seeing their confusion, knowing that there is no effective sanction for such behavior in an online community. The most effective countermove, by popular opinion, is to completely ignore the troll: “Do not feed the troll.” There also exists the unintentional troll, who is unaware that he or she is in fact trolling. If a troll is motivated by evil-minded pleasure and a secure feeling because there are no effective countermoves, it is not surprising that an intentional troll occurs. However, an unintentional troll is also not trivial. To understand a troll thoroughly, we formulated an anti-community strategy and a pro-community strategy, to express a troll’s behavior and a community member’s behavior, respectively, and we also assumed a third strategy: a de-community strategy. We executed evolutionary simulations with the three strategies, and then examined their outcomes. We found that we could replicate the situation where an excessive response against a troll leads to the closure of an online community and that an unintentional troll can have the same result. We also found the cyclical pattern of a dominant strategy such as rock-paper-scissors.
Yutaka Nakai

Economic and Social Networks


Chapter 5. On the Indeterminacy of the Clearing Payment Vectors in Numerical Simulations on Financial Networks

This paper points out a methodological lacuna in the recent stream of numerical analyses of contagion in financial networks, and presents a solution to amend it. Under some conditions, the intercyclical obligations that connect the agents in a financial network cause the indeterminacy of the vector of payments that clears such obligations. This problem, first pointed out by Eisenberg and Noe (Manage Sci 47:236–249, 2001), has received little or no attention by authors who investigate payment flows and domino effects in networks using numerical simulations. Here we present an original result that establishes necessary and sufficient conditions for the uniqueness of the clearing payment vector for any financial network, and we demonstrate this result to control for the occurrence of the above-mentioned indeterminacy while performing numerical exercises on financial networks.
Mario Eboli

Chapter 6. Three-State Opinion Formation Model on Adaptive Networks and Time to Consensus

We investigate the three-state majority rule model in a coevolving network with intensive average degree using Monte Carlo simulations. The key parameter investigated is the degree of homophily (heterophily), which is the probability p (\(q = 1 - p\)) of a given person being affected by others with the same (different) opinion. For a system with a uniformly random initial state, so that each person has an equal chance of selecting one of the three opinions, based on extensive Monte Carlo simulations, we found that there are three distinct phases: (1) When the population has an intermediate homophilic tendency, it reaches the consensus state very fast. (2) When the system has a moderate to large heterophilic tendency (a small value of p), the time to consensus (or convergence time) can be significantly longer. (3) When the system has a high homophilic tendency (a large value of p), the population can remain in a polarization state for a long time. We defined the convergence time for the system of voters to reach consensus operationally, and obtained a distribution function for the convergence time through Monte Carlo simulations. We observed that the average convergence time in a three-state opinion formation process is generally faster than in the same model of voting dynamics when only two states are available to voters, given that in the beginning of the simulations, different opinions are uniformly spread in the population. The implications in diversity are discussed.
Degang Wu, Kwok Yip Szeto

Chapter 7. Achieving Consensus with Segregation in Multiple Social Contexts

In social simulation, not only are the structures of the social relations fundamental for the construction of plausible scenarios, but also the interaction processes are shaped by these structures. Each actor interacts in multiple social contexts located within multiple social relations that constitute his social space. We build on previous work about context switching to study the notion of context segregation. The agents not only switch between social contexts, carrying with them their unique social identities, but also choose the contexts according to personal reasons. We apply the notion of context segregation to a simple game of consensus in which agents try to collectively achieve an essentially arbitrary consensus. We make a first analysis of our set of experiments towards understanding the influence of the segregation mechanism in the speed of convergence to global consensus and compare the results with those of the context switching model.
Davide Nunes, Luis Antunes

Behavioral Finance and Macroeconomics


Chapter 8. How Does Overconfidence Affect Asset Pricing, Volatility, and Volume?

Overconfidence is one of the most important characteristics of traders. In the past decade, theoretical approaches have paid much attention to this topic and obtained significant results. However, they heavily rely on specific assumptions regarding the characteristics of traders as well as the market environments. Most importantly, they only consider the market with a few types of traders. None of them is built upon a truly heterogeneous-agent framework. This paper develops an agent-based financial market. Each trader adopts a genetic programming learning method to form his expectations regarding the future. The overconfidence level of each trader is modeled as the degree of underestimation of the conditional variance. Based on this framework, we examine how traders’ overconfidence affects the market by analyzing the results regarding market volatility, price distortion, and trading volume.
Chia-Hsuan Yeh, Chun-Yi Yang

Chapter 9. Analyzing the Validity of Passive Investment Strategies Under Financial Constraints

This chapter describes the validity of a passive investment strategy through agent-based simulation. As a result of intensive experimentation, I have concluded that a passive investment strategy is valid under conditions where market prices deviate widely from fundamental values. However, my agent-based simulation also shows that the increase in the rate of passive investment slows as financial restrictions become more severe. The results are of both academic interest and practical use.
Hiroshi Takahashi

Chapter 10. Macroeconomic Forecasting with Agent-Based Models: Prediction and Simulation of the Impact of Public Policies on SMEs

The aim of this chapter is to show and discuss an integrated scheme of the MOSIPS project. This project, using an agent-based model methodology, focuses on simulating and evaluating policies for small and medium enterprises within a local environment. Its purpose is to conduct experiments on the implementation of policies according to different socio-economic scenarios. The results of these experiments allow stakeholders and citizens to know how the measures proposed by governments affect them and enable them to interact in the decision-making process. This perspective allows the analysis and implementation of the actions, interactions, and results between agents. At the micro-level, the model provides two main kinds of agents: individuals and firms, and some other entities: the public sector, the financial system, and the external sector. It also determines rules for actions and interactions. Simulations of variables and parameters lead to the analysis of the evolution and the forecast of the system at the macroeconomic level. The trajectories of the agents offer a dynamic standpoint that overcomes the standard equilibrium models.
Federico Pablo-Marti, Antonio Garcia-Tabuenca, Juan Luis Santos, Maria Teresa Gallo, Maria Teresa del Val, Tomas Mancha

Chapter 11. Influence of the Corporation Tax Rate on GDP in an Agent-Based Artificial Economic System

An agent-based model of an artificial economic system, including the government, was developed on the basis of the authors’ previous model. This model was used to analyze the factors influencing the relationship between GDP and the corporation tax rate and its mechanism. The findings show that executive compensation and the use of producers’ own cash for investment are both indispensable factors for reproducing the negative correlation between GDP and the corporation tax rate, because these actions help redistribute the firm’s surplus money to the market. Inefficiency in government expenditure is another indispensable condition. The calculated average multipliers for the reduction of both income tax and corporation tax are in good agreement with real data based on the macroeconometric model.
Shigeaki Ogibayashi, Kousei Takashima

Demographics, Health Care, Linguistics, and Sociology


Chapter 12. Semi-Artificial Models of Populations: Connecting Demography with Agent-Based Modelling

In this paper we present an agent-based model of the dynamics of mortality, fertility, and partnership formation in a closed population. Our goal is to bridge the methodological and conceptual gaps that remain between demography and agent-based social simulation approaches. The model construction incorporates elements of both perspectives, with demography contributing empirical data on population dynamics, subsequently embedded in an agent-based model situated on a 2D grid space. While taking inspiration from previous work applying agent-based simulation methodologies to demography, we extend this basic concept to a complete model of population change, which includes spatial elements as well as additional agent properties. Given the connection to empirical work based on demographic data for the United Kingdom, this model allows us to analyse population dynamics on several levels, from the individual, to the household, and to the whole simulated population. We propose that such an approach bolsters the strength of demographic analysis, adding additional explanatory power.
Eric Silverman, Jakub Bijak, Jason Noble, Viet Cao, Jason Hilton

Chapter 13. An Agent-Based Approach for Patient Satisfaction and Collateral Health Effects

The purpose of this study is to clarify the collateral health effects of health care, especially the relationship between patients and their families, using agent-based simulation. To this end we describe the general appearance of our simulation model and the simulation settings. The results of six model scenarios, each involving differing combinations of patient agents, patient’s family agents, doctor agents, a government agent, and nonprofit organization (NPO) agents, are then explained and discussed. We conclude with a summary that touches on the tasks that lie ahead, including an appropriate subset of health care policies that involve the participation of NPO agents.
Masatoshi Murakami, Noriyuki Tanida

Chapter 14. Complex Evolutionary Pathways in Interacting Linguistic Communities

We experiment with a linguistic change mechanism in a community of interacting agents and show the various phenomena that may emerge under different social constraints. We assume phonological and lexical learning and a semantic reference to external objects in the environment. Distinct groups of agents with initially different languages converge to a common language, with the relevant frequency of inter-agent interactions controlling which language dominates. Moreover, an initially monolingual community diverges due to social factors creating agent grouping, where agents interact more frequently with members of the same group. Additional cognitive features, like innovation and attention, lead to increased linguistic divergence between groups and word bistability. Finally, cultural learning leads to continuous linguistic change and occasional coexistence of multiple words, as well as revival of rare words. Overall, it appears that the initial community may evolve in arbitrary directions, and languages may dynamically form, split, mutate, and oscillate.
Ioannis Vagias, Elpida Tzafestas

Chapter 15. Socio-Cognitive Influences on Social Stratification

We study a simple social competition mechanism and the effect of various cognitive variants on the emerging stratification structure. We depart from a basic mammal-inspired competition mechanism relying solely on physical strength and introduce a number of higher level cognitive parameters, especially individual imposition and submission factors that modulate the agents’ decision to fight. The emerging structure is generally a society spread out in status range, without any obvious polarization. Next, the concept of a “stratification game” is introduced, where some fights are spontaneously resolved if differences between agents are mutually perceived as too large. Some game variants lead to fairly egalitarian societies, while others, based on withdrawal from fights, lead to extremely unequal societies. This property also holds in realistic competition environments involving resource sharing and task competition. Overall, our study shows how certain socio-cultural conventions may overtake cognitive factors and define the emerging social stratification outcome.
Elpida Tzafestas

Participatory Modeling


Chapter 16. A Computational Study of Rule Learning in “Do-It-Yourself Lottery” with Aggregate Information

This chapter computationally studies Barrow’s “do-it-yourself lottery” where players choose a positive integer that is expected to be the smallest one that is not chosen by anyone else. Here, we employ and modify the rule learning framework by Stahl (Games Econ Behav 32:105–138, 2000) based on the experimental findings by Östling et al. (Am Econ J Microecon 3:1–33, 2011), and incorporate them into our simulation model to see individual and collective behavior by changing the numbers of players and the upper limit. Our main conclusion is threefold: First, the game dynamics depends on both the number of players and the upper limit. Second, a lottery with a large sensitivity parameter divides the players into winner(s) and losers. Third, finding the “stick” rule immediately makes a player a winner and imitating behavior is not observed in four-player games.
Takashi Yamada, Takao Terano

Chapter 17. Agent-Based Social Simulation as an Aid to Communication Between Stakeholders

Various methods provided in conventional agent-based social simulation (ABSS) research are useful for modelers and analysts in evaluating its effectiveness. We know very little about how ABSS contributes to the decision-making process for practical business problems when used by managers and employees who are not familiar with it. In this research we talked to stakeholders in two complex and uncertain business situations about using the simulation results with our models. We found that ABSS helped to promote communication between stakeholders.
Kotaro Ohori, Shohei Yamane, Noriyuki Kobayashi, Akihiko Obata, Shingo Takahashi

Chapter 18. Hybrid Approach of Agent-Based and Gaming Simulations for Stakeholder Accreditation

We propose a hybrid approach of agent-based and gaming simulations for stakeholder accreditation. This approach assumes greater uncertainties about an agent’s decision-making and learning processes, severe time constraints on stakeholder participation in the modeling and simulation process, and limited understanding of an agent-based modeling language by stakeholders. The approach transforms an agent-based model into a game with adequate similarities between them, and it allows stakeholders to understand an agent-based model by playing the game. We developed a transformation-modeling protocol to convert an agent-based model into a valid and playable card game, and we introduce an example in which an agent-based model is transformed to assist the design of a performance measurement system in a sales organization. We report preliminary results for our approach.
Yusuke Goto, Yosuke Takizawa, Shingo Takahashi



Chapter 19. When Does Simulated Data Match Real Data?

Comparing Model Calibration Functions Using Genetic Algorithms
Agent-based models can be calibrated to replicate real-world data sets, but choosing the best set of parameters to achieve this result can be difficult. To validate a model, the real-world data set is often divided into a training and a test set. The training set is used to calibrate the parameters, and the test set is used to determine if the calibrated model represents the real-world data. The difference between the real-world data and the simulated data is determined using an error measure. When using evolutionary computation to choose the parameters, this error measure becomes the fitness function, and choosing the appropriate measure becomes even more crucial for a successful calibration process. We survey the effect of five different error measures in the context of a toy problem and a real-world problem (simulating online news consumption). We use each error measure in turn to calibrate on the training data set, and then examine the results of all five error measures on both the training and test data sets. For the toy problem, one measure was the Pareto-dominant choice for calibration, but no error measure dominated all the others for the real-world problem. Additionally, we observe the counterintuitive result that calibrating using one measure may sometimes lead to better performance on a second measure than could be achieved by calibrating using that second measure directly.
Forrest Stonedahl, William Rand

Chapter 20. Towards Validating a Model of Households and Societies in East Africa

One of the major challenges of social simulations is the validation of the models. When modeling societies, where experimentation is not practical or ethical, validation of models is inherently difficult. However, one of the significant strengths of the agent-based modeling (ABM) approach is that it begins with the implementation of a theory of behavior for relatively low-level agents and then produces high-level behaviors emerging from the low-level theory’s implementation. Our ABM model of societies is based on modeling the decision making of rural households in a 1,600 km (1,000 mile) square around Lake Victoria in East Africa. We report on the first validation of our model of households making their living on a daily basis by comparing resulting activities against societal data collected by anthropologists.
William G. Kennedy, Chenna Reddy Cotla, Tim Gulden, Mark Coletti, Claudio Cioffi-Revilla

Chapter 21. Social Simulation Comparison in Arbitrary Problem Domains: First Steps Towards a More Principled Approach

We outline a simulation development process, backed by a software framework, which focuses on developing and using a partial conceptual model as a ‘lens’ to compare and possibly re-implement existing models in a chosen problem domain (as well as to design new models). To make this feasible for existing models of arbitrary structure and background social theory, we construct our (partial) conceptual model in a way that acknowledges that it is a base representation which any individual model will typically add detail to, and abstract away from, in various ways which we argue can be formalised. A given model’s design is fitted to the conceptual model to capture how its structural architecture (and selected aspects of the system’s state and driving processes) map to the conceptual model. This fit can be used to produce incomplete skeleton code which can then be extended to produce a simulation. Along the way, we use robust decision-making to provide a useful frame and discuss how our approach differs from others. This is inevitably a preliminary approach to a broad and difficult problem, which we explore in the conclusions.
Stuart Rossiter, Jason Noble, Keith R. W. Bell


Weitere Informationen

Premium Partner