Introduction to Computational Social Science

What is Computational Social Science (CSS)? What are the main areas of this new, emerging field? What are the main assumptions and potential contributions of CSS? How does CSS differ from traditional social science disciplines? How does it differ from computer science? This chapter introduces the reader to the field of CSS, defined as the interdisciplinary conduct of social science research through an information-processing and complex adaptive systems paradigm, using computation as the key enabling scientific methodology. After exploring the definition of CSS and the computational paradigm of society, the chapter provides examples of CSS investigations in basic and applied domains across the social sciences and areas of policy analysis. The concept of a complex adaptive system is introduced in the context of Herbert A. Simon's fundamental theory of artifacts, especially in terms of explaining the origin and development of social complexity and civilization—both ancient and contemporary. An overview of the main areas of CSS is provided, including computational content analysis, social networks, social complexity, and social simulation models. Each area of CSS is detailed in subsequent chapters. The chapter concludes with a historical overview of CSS to explain the scientific roots and main trends of the field.

Social scientists have used computation since the days of the earliest digital computers. What is the role of computation in contemporary Computational Social Science (CSS) theory and research? How does computation provide a deeper understanding of social complexity? This chapter is not an introduction to computing for social scientists. Rather, it is an examination of computation from a CSS perspective; similar to how a computational astronomer or a computational biologist would discuss the function of computation in their respective disciplines. After examining key similarities and differences between computers and social systems, from an information-processing perspective, the chapter takes a closer look at programming languages and aspects of implementation. Classes, objects, and dynamics are examined from the perspective of social theories and the role computational entities play in the conduct of CSS research. The Unified Modeling Language (UML) is used as a systematic graphic notation for representing social entities, relations, and interactions, based on a variety of examples drawn from across social science domains. Data structures and algorithms, which are foundational to computation, are examined in the context of CSS research.

Chapter 1 identified automated information extraction (also known as computational content analysis or media-mining) as the first area of Computational Social Science. Chapter 3 takes a close look at this area, beginning with roots in linguistics. Computational text mining has been the main application of this area of CSS, but audio, imagery, and social media data are also components of the expanding Big Data universe. Theory and research in automated information extraction is at the base of major social science discoveries, such as universal semantic spaces and the fundamental structure of human information-processing. A major focus of this chapter is on the methodology of automated information extraction, including phases that extend from the formulation of research questions to the selection of sources, preprocessing preparations, to analysis in a technical sense. Illustrative examples are provided, including some recent transformative breakthroughs in computational events data analysis and geospatial data structures. The material in this chapter has intrinsic value as well as being instrumental for understanding networks, complexity, and simulation modeling approaches in subsequent chapters.

The concept of a network is foundational to social science in general and to CSS in particular. How does CSS investigate networks across domains of social systems and processes? What do computational approaches add to the development of theory and research in social networks science? How can CSS deepen our understanding of social networks? This chapter introduces the reader to elements of social network analysis and computational applications to analyzing social complexity. The sections of this chapter cover formal aspects that have universal application to many different kinds of social networks, as well as applications to significant areas such as human cognition and decision-making, organizational models, the structure of small worlds, and international relations. From a formal perspective, the chapter highlights both mathematical and computational aspects of social networks. From a CSS perspective, social networks can be constructed via automated information extraction algorithms, based on ideas from Chap. 3. Social networks can also be used as a basis for analyzing social complexity and simulation models, as in the next chapters.

Social complexity is a fundamental concept in Computational Social Science (CSS), based on Simon's theory of artifacts and the complex adaptive systems paradigm. This is the first of three chapters on social complexity. The emphasis in this chapter is primarily on descriptive aspects. After providing some initial working definitions of social complexity—which are further developed in the next two chapters—the chapter examines empirical, descriptive aspects of where, when, and how social complexity originated in the “cradles of civilization.” This material on descriptive origins of social complexity is critical for a factual understanding of social complexity as a global, cross-cultural phenomenon. Important conceptual aspects of social complexity also include a closer examination of bounded rationality and near-decomposability, ideas central to CSS theory and research. The social complexity that exists today originated thousands of years ago at specific locations under specific circumstances, not in some arbitrary way. How do we know this from a scientific perspective? This chapter addresses this question by examining current methods for measuring social complexity, including the use of multiple lines of evidence and quantitative scales of complexity and related phenomena.

Laws describe; theories explain. This chapter introduces the reader to laws of social complexity. These are formal descriptive relations that exist among components and variables of complex social systems. The approach in this chapter is to divide the study of laws of social complexity into two classes: structural laws and distributional laws. Structural laws of social complexity describe how social complexity depends on organizational architecture. The two fundamental patterns of structural architecture in structural laws of social complexity are serial and parallel, which correspond to conjunctive and disjunctive events in the performance of social systems and processes. The most complex social systems are ultimately based on these fundamental patterns. By contrast, distributional laws of social complexity describe how social complexity depends on non-equilibrium dynamics. The fundamental pattern of non-equilibrium dynamics in social complexity is the power law, which assumes various forms in different social systems and processes. Some of the most intriguing complex social phenomena (language, conflict, wealth, urbanization) are based on power laws and non-equilibrium dynamics.

Since laws describe and theories explain, this chapter introduces the reader to generative theories of social complexity in CSS. The emphasis is on formal and empirically-referenced theories, in some cases amenable to validation with real-world data, consistent with a computational approach. The chapter has three thrusts. The first is to provide the reader with proper foundations in elements of process analysis, Boolean logic, and elementary probability for compound events, as these ideas are applied to the statement and development of social complexity theories. This is done by highlighting useful parallels across formalisms and making use of mathematical isomorphisms. The second thrust of this chapter is to use these ideas as building blocks to examine theories of origins of social complexity. This is done by taking a closer look at how and why chiefdoms—and, later, states and empires—formed in the first cradles of civilization. Detailed analysis of these primary systems and processes is essential for understanding social complexity. The third thrust of the chapter is on general theories of social complexity, which encompass universal explanations of original, current, and future social complexity. The theoretical concepts, principles, and methods covered in this chapter constitute one of the most dynamic frontiers of CSS.

Social simulation is the fourth major area of Computational Social Science (CSS). Social simulations are used in CSS as artificial worlds for conducting virtual experiments, analyzing scenarios, and exploring alternative pasts and futures. In a sense, social simulations are spatio-temporal machines used by CSS researchers to explore the social universe in unique ways that are not accessible through other scientific instruments. This chapter introduces the reader to the methodology of social simulations. This is done in terms of a general “life cycle” framework that begins with research questions and develops through stages of modeling, implementation, verification, validation, analysis, and communication. This process is cyclical, not linear, although the exposition of the process is necessarily sequential. In practice, feedback from consecutive stages of simulation model development is used to improve previous and subsequent stages. What matters most is maintaining quality control throughout the process in order to be able to analyze a model that can be trusted. Complex social simulations are a breed apart, comprised of relatively large teams consisting of multi-disciplinary specialists, multi-institutional arrangements, and multi-year duration. The chapter also presents several major types of simulations, including variable-oriented models and object-oriented models, which are examined in detail in the next two chapters.

Social simulations became famous in the 1960s through dynamical models used to predict the future of the world in terms of population growth, urbanization, pollution, and other variables. This chapter examines the contemporary landscape of variable-based model used in CSS. The emphasis is on system dynamics models and queuing models, two of the most important classes of variable-oriented models in use today. Both social simulation modeling traditions are examined in terms of the methodology introduced in the previous chapter and examples are provided from areas of pure and applied CSS research. In general, system dynamics simulations emphasize deterministic causal processes, whereas queuing simulations highlight stochastic processes and mean, statistical behaviors. However, this is only a matter of general orientation, since both classes of models can share some features of the other, especially system dynamics models with probabilistic components. Both classes of models comprise a large scientific literature and numerous applications that range from basic scientific research to applied policy domains in national defense and homeland security, transportation, public health, commercial and management applications, to name just a few.

This chapter examines the newest types of social simulations, which are object-based to represent social entities in ways that are more explicit than through variable-oriented models. The social simulations examined in this chapter constitute a rich and growing family of generative models consisting of two main classes: cellular automata and agent-based or multi-agent models. Both are generally seen as spatial models, but they can as easily capture network aspects of social complexity. Processes of urbanization and opinion dynamics are often simulated through the use of cellular automata. Spatial or organizational agent-based models are used to simulate social complexity in an increasing variety of domains, ranging from cultural dynamics to financial crises; from regional transportation systems to public health management; from humanitarian crises to global processes, such as climate change and the rise and fall of polities and civilizations. Agent-based simulations do this by using the versatility of multi-agent systems as computational frameworks within which modelers and programmers integrate social, natural, and artificial entities and dynamics. These social simulations often make use of other methodologies, such as network analysis or geographic information systems and remote sensing data.