CMC

SNA in Computer-Mediated Communication

Computer-Mediated Communication (CMC) has rapidly evolved into a primary form of communication in a broad range of interpersonal, organizational, and mass communication functions. As Internet technologies have developed, forms of computer-mediated communication have proliferated. Among the more obvious forms are email, discussion groups, Usenet newsgroups, Internet Relay Chat, Instant Messaging, and real-time audio and video chat. Weblogs and online multiplayer games represent forms of CMC that have more recently acquired popularity. Understanding CMC is now a key enterprise of the social sciences and humanities, with participation from diverse fields such as psychology, sociology, linguistics, anthropology, and rhetoric, among others, in addition to the applied fields of human-computer interaction and computer-supported cooperative work.

Among the things we most urgently need to understand are the large-scale patterns of CMC use, whether these involve technology adoption, information diffusion, the creation of new forms and genres of communication, or more complex social dynamics. Patterns of use change with each new generation of Internet users and/or Internet applications. The early Internet was characterized by a small number of mostly elite (academic and research-institution based) users employing primarily textual modes of communication, today's uses employ a variety of rich multimedia features such as graphics, animation, video and audio. Moreover, whereas the social context of communication was largely implicit in earlier forms of CMC, current modes encourage users to directly symbolize their ties to other users in the form of friends lists, buddy lists and other similar mechanisms.

As the technical environment changes, so too does the social environment. The hundreds of millions of Internet users today are more geographically, linguistically, and culturally diverse than the early adopter pool. The users of even a single service such as LiveJournal or MySpace number in the tens of millions -- typically more than the entire Internet when it first became a global force.

Social Network Analysis offers an important perspective on Intenet CMC use because it allows one to directly connect observation of individual-level actions with global trends. It is applicable to a broad range of questions about information and communication technologies and their social effects, and permits us to compare the effects of different CMC types, to each other and to face-to-face communication. On the theoretical level, SNA is especially useful in developing understandings of the social dynamics of CMC use.

Social Network Analysis

Social Network Analysis has three primary components:

  1. The application of graph theory, a branch of mathematics, to the analysis of social relations
  2. The development and application of statistical models applied to graphs
  3. The development of social interpretations of observations and generalizations obtained through the above two

Graph Theory

Graph theory uses set-theoretic constructs to describe entities and their basic and derived relations to one another. The entities are nodes in the graph, while (basic) relations are represented as arcs if they are directed (asymmetric) relations, or edges if they are undirected (symmetric) relations. For social network analysis, the nodes are actors of some sort, usually individuals or groups of individuals. Relations considered are typically social relations like "friend of", "acquaintance of", "exchanges goods with", etc.

Statistical Models

Two general types of statistical models are used in the analysis of graphs: Latent Space models, and Generalized Linear Models. Latent Space models, such as Principal Components Analysis, Factor Analysis and related models, are used to simplify the representation of social relations so that socailly similar actors — actors who have more or less similar social relations to other actors — are represented as occupying similar locations in an abstract space. There are many methods for doing this, and many times latent space analysis is followed by cluster analysis in order to identify classes of equivalent actors. The social construct that this form of analysis is most closely associated with is structural equivalence, wherein actors are regarded as equivalent only when they have similar ties to the same other actors. Regular equivalence, in which actors are equivalent if they have similar ties to some other actors, is not well addressed using these methods. Since regular equivalence is often the prefered notion for many questions, alternative models (e.g. Generalized Blockmodeling, Doreian, et al.) have been proposed.

The approach known as Exponential Random Graph Modeling (ERGM) or p* (pronounced "P-star") modeling represent the application of Generalized Linear Models (GLMs) to graphs. If relations are binary valued, logit/logistic regression models are typically used (though other binomial models can be used as well). When relations are treated as integer-valued, general log-linear models may be used. Sometimes hybrid conditional models are used if the values involved are distributed in some more complex way.

The ERGM approach has a couple of less obvious reasons to recommend it. First, unlike other modeling approaches, it is possible to account for multiple relation types within a single model in a somewhat more natural way. The relation type is represented as an independent variable in the model, which may be involved in interactions like other independent variables. Differences among relations, where justified by significance testing, and be addressed through such interactions. Second, it is possible to represent notions such as different types of regular equivalence in a similar manner. In this way, the model with a very general character, and with well understood statistical properties.

Interpretation

Interpretation in SNA focuses on four general areas:

  1. Connection
  2. Structure
  3. Flows and exchanges
  4. Network Dynamics

Connection is the most accessible of these, and perhaps the least interesting from a social perspective. The basic idea is that people may be more or less connected individually, and hence have greater or lesser access to social resources, and that overall people are relatively more closely connected than might be naively expected, what is known as the small-world phenomenon. The most often cited result here is Stanley Milgram's "six degrees of separation" between two arbitrary people, from the famous Travers and Milgram (1969) study, which is unfortunately often mis-cited and exaggerated. People interested in this idea may pursue the original paper and subsequent works.

Social structure is one of the central concepts to the study of SNA. Notions of equivalence (structural and regular), social roles, centrality and power are all structural in nature, and it is through excavating and laying bare the social structure of different interactional contexts that we get much of the impact of SNA on the analysis of CMC. Appreciating the structure of interaction around different communicative contexts allows us to develop rich understandings of how some participants become empowered while others are disempowered, how participants are valued and how social systems are directed, whether according to or in spite of their overt institutional structures.

Structure leads naturally to consideration of flows and exchanges, where ties either condition or represent flows of material or symbolic value. Electronic commerce is most clearly relevant to notions of flow where the basic nature of the phenomenon is both economic and CMC-based. But other kinds of flow are crucial in the context of CMC as well. For example, the notion of information flow is directly implicated, and other kinds of flows, such as the adoption of innovations or technologies can be viewed in the same way. Flows may be more or less balanced, and we may speak of the accumulation of wealth, symbolic capital, power, knowledge etc. on the one hand, as opposed to more mutualistic relations such as support on the other.

Finally, it is necessary to recognize that social relations themselves are not static, but rather are dynamic in character. Members enter and leave a network, ties are formed and/or broken, roles pass from one actor to another, etc. Full understanding of a network context is possible only when we understand how connection, structure, and flows unfold and develop over time. Dynamic models of social networks are more difficult to construct, but highly rewarding in the richness of interpretation that they offer.


Categories: Research, SNA, Statistics, Interpretation