Social Networks

An Introduction

Jeroen Bruggeman

• Publisher: Routledge
• Year: 2013
• eBook ISBN: 9781134048199
• Topic: Social Sciences
• Subtopic: Sociology
• Index: Social Sciences

Social Networks: An Introduction is the first textbook that combines new with still-valuable older methods and theories.

Designed to be a core text for graduate (and some undergraduate) courses in a variety of disciplines it is well-suited for everybody who makes a first encounter with the field of social networks, both academics and practitioners.

This book includes reviews, study questions and text boxes as well as using innovative pedagogy to explain mathematical models and concepts. Examples ranging from anthropology to organizational sociology and business studies ensure wide applicability. An easy to use software tool, free of charge and open source, is appended on the supporting website that enables readers to depict and analyze networks of their interest.

It is essential reading for students in sociology, anthropology, and business studies and can be used as secondary material for courses in economics and political science.

Chapter 1

Introduction

Social Relations

Humans are more cooperative with non-kin and exchange more information than all other animals, with large ramifications for their societies, which eventually became a globe-spanning network. Their pro-sociality enables them to specialize in few activities, and renders them dependent on others for their remaining needs and desires to be fulfilled (Smith 1986). Consequently, one might think of individuals as free-floating atoms in markets where they exchange goods and services. On closer inspection, however, humans rather prefer to affiliate themselves with others in groups and communities of all sorts, such as families, settlements, religions, organizations, and sometimes virtual communities as well. Even if they engage in seemingly free market transactions, these transactions are embedded in institutions beyond bilateral relations, e.g. property rights (Granovetter 1985; North 1990). As part of their gregariousness, humans also learn most of what they know from others, including from the media these others may use. Humans are able collectively to support each other and to invent and produce on an unprecedented scale, but also to harm each other and their natural environment (Tinbergen 1968). For better and for worse, people are profoundly influenced by others for most of what they have, know, and do. These interdependences imply neither that people react to information or influence in a uniform way nor that everybody becomes similar. They do imply, however, that, when we attempt to comprehend humans, studying them in their social environment is much more illuminating than seeing them only as individuals. We should therefore focus on social relations in the most general sense (Emirbayer 1997; Elias 1970). Without studying these relations, social phenomena such as religious practices, collective violence, technological innovations, searching resources, and tastes in music and fashion would not be comprehensible, since they cannot be reduced to properties of individuals (Hedström 2006), just as music cannot be appreciated if it is perceived as a series of individual notes. Therefore only the study of social relations can help to reveal the mechanisms that determine social developments, which in their turn set the conditions for continuing, interrupting, and newly establishing relations, as well as the personal experiences these relations bring about.¹ The meanings assigned to these experiences, in turn, are also of interest to social workers, consultants, and other practitioners.

Relations Represented as Networks: Overview

Comprehending the raw complexity of social relations beyond a very small number of people is not feasible without conceptual tools. In this book we will see how social relations can be systematically and clearly represented as networks. A network can depict in a single graph a multitude of interactions between many people, which might have taken place at different times and places. In the words of Jeremy Boissevain (1979), “Network analysis asks questions about who is linked to whom, the content of the linkages, the pattern they form, the relation between the pattern and behaviour, and the relation between the pattern and other social factors.” From a network perspective we will investigate human society. We will see that there exist patterns of relations that are crucial for the flows of information, influence, goods, and contagious diseases, that hardly anyone could imagine before they were discovered. We will focus in particular on communities in the broadest sense, because they lie at the core of human sociality. Communities and their overlaps can be detected on the basis of relational data only, without information about individuals’ attributes. A case in point is a group partitioned into two opposing coalitions by conflict. We will also see how a community’s social cohesion can be measured. From a cultural-evolutionary perspective we will then discuss why people engage extensively in communities and in the maintenance of seemingly inefficient social relations, in contrast to living by arm’s-length market relations alone. Then we will shift our focus from communities to individuals and see how people can get access to resources by being connected to others. People as community members can at the same time collaborate for collective goals, while their network positions enable them to broker information and to achieve power and status, leading to social inequality. In these network positions they have certain roles, seen as typical patterns of relations, which can be assessed precisely. Our focus on individuals will also point out why new culture is most likely to be created at the interstices of (sub) communities, where information brokers reside. Newly created knowledge can diffuse through communities by several mechanisms of transmission that we will point out. Organizations, which may be regarded as “special-purpose” communities, are treated at greater length than other kinds of community because many of us work in or for them, and everybody has to deal with them sometimes. Finally, we will show how to use the computer to analyze social networks, and will provide guidelines for data collection. For those who are new to networks, the approach taken might be an unfamiliar but hopefully illuminating way to look at familiar social and sociological problems.

The field of social network analysis up to the mid 1990s was covered extensively in a voluminous handbook by Stanley Wasserman and Katherine Faust (1994). Then two papers by the physicists Duncan Watts and Steven Strogatz (1998) and Albert-László Barabási and Réka Albert (1999) revolutionized the field, and in their trail a highly interdisciplinary and rapidly growing literature transformed it into a broader science of complex networks (Boccaletti et al. 2006). Meanwhile, the sociological approach evolved in its own way (Carrington, Scott, and Wasserman 2005). In this book, one will find a selection from both streams of literature and from their cross-fertilizations. The purpose of this book is introductory, and I chose those subjects that most people deal with, rather than attempting to be comprehensive in my overview of substantive research areas where network analysis is applied, or of models it has to offer. My selection is organized around sociological rather than network subjects, a subtle but important distinction with handbooks, while acknowledging that developments in the field of networks shape our perceptions and categorizations of the subject matter. As a consequence of my focus, this book is as much an introduction to social networks as it is an (advanced) introduction to sociology from a network perspective. I assumed that readers might first want to see what network analysis has to offer before they would get interested in its history, and therefore left out the latter. Just in case, there is a short review by Charles Kadushin (2005) of Linton Freeman’s (2004) book-length history, while the science of complex networks is described by Barabási (2002).

Other Approaches

Social network analysis is not the only approach that focuses on interdependent actors at the micro level and the consequences of their behavior at the macro level. Of other approaches that do so, game theory (Szabó and Fáth 2007) is probably the most well known; it analyzes how macro outcomes of individual decisions, based on self-interested motives, result from the interdependent decisions of all players in a game – another way to model social situations. Social network analysis has a different but complementary focus on patterns of relations, and one could say, paraphrasing the title of Thomas Schelling’s classic (1978), that it deals with micro motifs and macro behavior.² In general, the network approach complements other approaches to social phenomena in its systematic treatment of social relations, e.g. in sociology (H. White 1992), anthropology (D. White and Johansen 2005), history (Bearman, Moody, and Faris 2003), economics (Goyal 2007), social psychology (Moreno 1934), communication (Monge and Contractor 2003), political science (Mutz 2002), and organization science (Powell 1990). These fields in their turn provide valuable ideas and data for network analysis, as well as contexts for interpretation of results. For models and methods, the field of social networks traditionally cross-fertilizes with graph theory (Harary 1969), and more recently with physics (Strogatz 2001), biology (Ravasz et al. 2002), statistics (Snijders et al. 2006), linguistics (WordNet),³ engineering (Albert, Jeong, and Barabási 2000), and computer science (Dorogovtsev and Mendes 2002), blurring the boundaries between these fields.⁴

Selection Criteria

To choose material for this book from the vast literature, I applied four criteria that go beyond recent developments of various subfields: testability, clarity, parsimony, and relevance. First, human ideas are fallible, and science distinguishes itself from other fields by requiring that ideas be empirically testable, or yield testable predictions, as Willard Van Orman Quine (1990) would say. Testability extends to data and software, which should be open to inspection by other researchers. Obviously, this makes sense only if a falsehood or mistake exposed receives proper recognition. Sifting true from false scientific ideas is now taken for granted, and the struggles to get this practice accepted are long forgotten. However, it is important to realize that this was a great achievement, which can be recognized if we notice how hard it is to get rid of false beliefs in other realms of society, e.g. ethnic prejudice. We may thus acknowledge testability, but many ideas, even scientific ones, are not testable straight away. Scientific concepts, for instance social cohesion, usually have multiple meanings, some of which are more useful than others – if they can be made clear. Without guidance on which interpretations out of many make sense, theory testing and knowledge accumulation are not feasible and, on the basis of unclear concepts, computations – essential for grasping complexity – are impossible. In network research, as in many other areas of science, ideas can be made clear by expressing them in a formal language (Peirce 1878; Suppes 1968). Mathematics is used to explicate ideas and intuition in a concise manner, and may therefore be regarded as “the poetry of science” (De Swaan 1995). It is also used to model complex phenomena in a precise manner, and through computations it enables us to reach conclusions that would be too complex for unaided reason (Farquharson 1969). This book, however, can be read and understood without math, which has been put in text boxes and footnotes. Either way, actual computations we will do by computer.

No matter how much computer power is applied, however, science is a human activity and its results should be made understandable for humans as much as possible. Scientific theories should therefore be parsimonious. Before the digital age, theories were required to be outright simple (Quine and Ullian 1978), whereas nowadays one might say that the “computational complexity” of theories should be amenable for the human brain to digest. Digestibility depends partly on the amount of prior training of the brain, and therefore some network theories and models that require more mathematical background than can be laid out here are treated at an elementary level. In these cases, however, the basic principles are still simple and can be understood by non-experts, although the subsequent computations or algorithms based on these principles may be complex.

Along with simplicity, parsimony has a second and equally important meaning: that more phenomena can be explained by less theory, or captured by fewer models – generality for short. For example, the concept of hierarchy is very general because it applies to a great many phenomena, such as language structures, and rank orders in business firms, groups of apes, and Web directories. If parsimony were to be carried out to its extreme, however, it would yield an empty theory, perfectly simple and applicable to everything. Obviously, the empty theory provides no comprehension at all, and parsimony should be traded off with other requirements with which it is meshed (Goodman 1961). When shifting our focus to the phenomena to be understood, we may realize that not only their general patterns matter but also their variation. As linguists are interested not just in universal language structures but also in descriptions of different languages and their evolution, social scientists know that formal network models are often preceded, complemented, or otherwise enriched by ethnographic and historical studies. Comparing different variants of a phenomenon, often qualitatively described, can help us considerably to generalize. In a similar vein, applicants of statistical models have to trade off simplicity with more variation explained by more complex models. Furthermore, formalization is an enduring process, unfinished for many concepts, and some phenomena not yet properly cast in a precise language, e.g. culture, are too important to leave undiscussed. In general we should regard all approaches that stretch our imagination or increase our comprehension as relevant. In sum, when selecting scientific contributions or developing our own, we should aim at clarity, testability, relevance, and reasonable parsimony. These four criteria do not nearly exhaust all good lessons from philosophy of science but, as it seems, they cover the most important and most arduous problems in scientific practice, or of thought in general, as one might argue. To wrap up with the words of the Nobel laureate Herbert Simon (1996: x), “The goal of science is to make the wonderful and complex understandable and simple – but not less wonderful.” On this note, we may get started with social networks.

Suggested Background Reading

General history of mankind, Diamond (1997, 2002); introduction to social science, De Swaan (2001); embeddedness of social action in relations, Granovetter (1985); relational thinking in sociology, Emirbayer (1997); a brief historical overview of the field of network analysis, Kadushin (2005); an introduction to philosophy of science, Quine and Ullian (1978); finally, a delightful book on the beauty of mathematics is Aigner and Ziegler (2003), which suggests a fifth criterion that we might want to add to the four above.

Chapter 2

Representation and Conceptualization

Networks as Representations of Social Relations

Social relations exist in a large variety. Criminal contacts are kept secret, whereas presidential encounters are conspicuously shown on television for millions to watch. Friendships are mostly between (approximate) equals, whereas employment relations are authoritarian, and in some places involve force, at an opposite extreme from equality. Lovers have mutual trust, emotional involvement, and frequent interactions, whereas “arms length” market transactions are brief and emotionally shallow. Although many relations are reciprocated, they are not necessarily so, like fleeting glances in a subway. The last example shows that some relations are not even noticed by all participants but can still be influential: someone saw somebody with an appealing product and now wants to have it too. In strength, relations vary as well, often associated with the kind of relation. Strength varies as a function of (1) emotional intensity, (2) trust, (3) time spent, and (4) reciprocity (Granovetter 1973).¹ The more two people are emotionally involved with each other, trust each other, spend time on or with each other, and reciprocate each other’s actions, the stronger their relationship will be. However, the stronger relationships are, the more ambivalent people tend to feel about each other (Smelser 1998), which can sometimes result in conflict.

Graphs

In searching for relational patterns to explain social phenomena, as motivated by Chapter 1, it seems that we have gone astray, by tumbling into a bewildering variety of relations seemingly out of control. Continuing to study what we can hardly grasp seems not to be a very useful pursuit. What we need instead is a clear-cut representation or model of these social relations, in order to see what we are talking about. Subsequently, we need tools and theory² to analyze these relations, in order to develop a meaningful understanding. On both counts, social network analysis can help us to get ahead. Actors, i.e. intentional objects such as humans or organizations, represented as nodes (vertices) and their relations as lines (edges), can be drawn as a graph, or social network. The problem of getting empirical data to draw such a graph can then be treated separately. The resulting graph makes possible visual inspection of relations beyond anything possible by observing the actual relations, and a graph can be a great deal more complete and less ambiguous than the ephemeral encounters on which it is based.

As an example of a social network, let us look at a pattern of sexual relations (Figure 2.1). The da...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright
• Contents
• List of illustrations
• Acknowledgements
• 1. Introduction
• 2. Representation and conceptualization
• 3. Small worlds
• 4. Searching and fat tails
• 5. Communities: detection, conflict, cohesion, and culture
• 6. Social inequality: prestige, power, brokerage, and roles
• 7. Organizations as networks
• 8. Methods: data and software
• Glossary of network notions
• Notes
• Bibliography
• Index