Global Village or Cyber-Balkans? Modeling and Measuring the Integration of Electronic Communities

Posted: August 20, 2008 in 2005, Articles
Tags: ,

Marshall W. Van Alstyne
Boston University – Department of Management Information Systems; MIT – Center for E-Business

Erik Brynjolfsson
Massachusetts Institute of Technology (MIT) – Sloan School of Management; National Bureau of Economic Research (NBER)

ABSTRACT:

Information technology can link geographically separated people and help them locate interesting or useful resources. These attributes have the potential to bridge gaps and unite communities. Paradoxically, they also have the potential to fragment interaction and divide groups. Advances in technology can make it easier for people to spend more time on special interests and to screen out unwanted contact. Geographic boundaries can thus be supplanted by boundaries on other dimensions. This paper formally defines a precise set of measures of information integration and develops a model of individual knowledge profiles and community affiliation. These factors suggest specific conditions under which improved access, search, and screening can either integrate or fragment interaction on various dimensions. As IT capabilities continue to improve, preferences—not geography or technology—become the key determinants of community boundaries.

DOWNLOAD PDF (Click “Choose Download Location”)

SNIPPET:

Empowered by search engines, recommender systems, search agents, and automatic filters, information technology (IT) users are spending more of their waking hours on the Internet, choosing to interact with information sources customized to their individual interests. But, does the emergence of global information infrastructure necessarily imply the emergence of the global village–a virtual community of neighbors freed of geographic constraints? Or, will the borders merely shift from those based on geography to those based on interest?

In this paper, we show that an emerging global village represents only one of a range of possible outcomes. Improved communications access and filtering technologies can, in some circumstances, lead to more fragmented intellectual and social interaction. In particular, we show that preferences can reshape social, intellectual, and economic neighborhoods as distinct from those based on geography. Just as separation in physical space can divide geographic groups, we find that separation in virtual knowledge space can divide special interest groups. In certain cases, the latter can be more insular. We introduce several formal indices of integration and then show both algebraically and graphically the conditions under which these indices rise or fall with different preferences and levels of access.

The conclusion that increased connectivity and improved filtering can actually lead to less integration is based on two observations. First, bounded rationality–a limit on the human capacity for information processing (Simon 1957)–can lead to specialization, which decreases the range of overlapping activities. As IT eliminates geographical constraints on interaction, the constraints of bounded rationality become increasingly important. Information transmission and bandwidth have increased across all distances except the last 12 inches–between people and machines. Regardless of how fast data scrolls across the screen, absorption is bounded. In the limit, people must choose some information contacts over others. Filters, even sophisticated electronic filters, must be selective in order to provide value. Thus, certain contacts, ideas, or both, will be screened out.

The second observation is that IT can provide a lubricant that enables the satisfaction of preferences against the friction of geography. On the one hand, those with a preference for specialization, whether intrinsic or driven by external rewards, may seek more focused contact than available locally. Thus, local heterogeneity can give way to virtual homogeneity as specialized communities coalesce across geographic boundaries. On the other hand, preferences for broader knowledge, or even randomized information, can also be indulged. In the presence of IT, a taste for diverse interaction leads to greater integration–underscoring how the technology serves mainly to amplify individual preferences. IT does not predetermine one outcome.

The same mechanisms that affect the specialization of knowledge also affect the degree to which interactions among people and communities become more or less integrated. The Internet can provide access to millions of other users and a wide range of knowledge sources, but no one can interact with all of them. Bounded rationality implies that a citizen of cyberspace still has a finite set of “neighbors” with whom he or she can meaningfully interact, but that nongeographic criteria increasingly influence the selection of these neighbors. Nongeographic criteria for selecting acquaintances can include common interests, status, economic class, academic discipline, religion, politics, or ethnic group. In some cases, the result can be a greater balkanization along dimensions that matter far more than geography, while in other cases more diverse communities can emerge. Our analysis suggests that automatic search tools and filters that route communications among people based on their views, reputations, past statements, or personal characteristics are not necessarily benign in their effects.

Preferences themselves need not remain unaffected by such tools. Because the Internet makes it easier to find like-minded individuals, it can facilitate the creation and strength of fringe communities that have a common ideology but are dispersed geographically. Thus, particle physicists, oenophiles, Star Trek fans, and members of militia groups have used the Internet to find each other, swap information, and stoke each others’ passions. In many cases, their heated dialogues might never have reached critical mass as long as geographic separation diluted them to a few participants per million. Once connected, their subsequent interactions can further polarize their views or even ignite calls-to-action (Sunstein 2002). The Internet can also facilitate the de facto secession of individuals or groups from their geographic neighborhoods. One study found that increased hours spent using the Internet can be strongly associated with a loss of contact with one’s social environment and spending less time with human beings (Nie and Erbring 2000). Another study found that users decreased their local knowledge while their knowledge of national events remained about the same (Kraut et al. 2002). Consistent with the predisposition arguments presented below, the latter study also found that introverts decreased on measures of community involvement and increased in loneliness, while extroverts increased their involvement and decreased in loneliness. The Internet can apparently lead to spending less time interacting with geographic neighbors, isolating individuals on some dimensions even as it integrates them on others.

We do not argue that increased specialization or balkanization must always result from increased connectivity. On the contrary, we believe that the Internet has enormous potential to elevate the nature of human interaction. Indeed, we find that if preferences favor diversity, increased connectivity reduces specialization and increases integration. Strong ties and social bonding provide important social benefits (Wellman and Wortley 1990, Putnam 2000). However, our analysis also indicates, other factors being equal, all that is required to reduce integration in most cases is that preferred interactions are more focused than existing interactions. A desire for increased focus and improved filtering of noisy communications is a natural response to data and computational overload. Although the conventional wisdom has stressed the integrating effects of the technology, we examine critically the claim that a global village is the inexorable result of increased connectivity and develop a suite of formal measures to address this question.

2. Related Literature

To characterize group information sharing, we draw on related literature from a variety of perspectives, including theories of attraction (Blau 1977), dynamic social interaction (Latane 1996), group stability (Carley 1990), group diversity (Ancona and Caldwell 1992), social networks (White et al. 1976, Wellman and Wortley 1990, Wellman and Gulia 1997), network measures (Banks and Carley 1996, Sunil et al. 1995, Teachman 1980, Wasserman and Faust 1994, Watts and Strogatz 1998), and diffusion models (Valente 1995).

Like Blau (1977), we use an attribute vector, such as age, sex, race, religion, and employment, to predict social differentiation, group formation, and individual tendencies toward social interaction, but we focus on information access. Blau’s homophily model of attributes, for example, predicts that two white male postal workers share more in common than either might share with a black female executive. Based on differences among individuals and the assumption that influence declines with distance, Latane (1996) argues that group patterns emerge as a function of the strength, immediacy, and number of social factors acting on individuals. Latane’s Dynamic Social Impact Theory holds that people become more similar to their neighbors, leading to spatial clustering, and that changing patterns may exhibit nonlinearity as opinions resist outside pressure up to a threshold, which we model explicitly in Corollary 1. An empirical study in support of this theory found that group members came to resemble their neighbors in electronic space, opinions on unrelated topics became correlated, and majority factions increased in size, but minority factions became more coherent (Latane and Bourgeois 1996).

Group stability is also considered in Carley’s (1990, p. 332) “constructural” model, where groups “form and endure because of discrepancies in who knows what.” Shared knowledge leads to interaction and, in turn, interaction leads to shared knowledge. The modeling parameters and analysis resemble those introduced here, with a few exceptions. First, Carley’s simulation analysis tracks the complex dynamic character of group boundaries over time. In contrast, our derivations are analytical and focus on comparative static results and equilibrium conditions. Second, most models of this type (e.g., triad completion, constructural, degree variance) eventually homogenize in the sense that interaction probabilities between all pairs of agents become equal (Banks and Carley 1996). In our model, homogenization and balkanization can both result. The key difference is the interaction of preferences with bounded capacity; for if agents in our model had unbounded capacity, integration would always result. Indeed, even with bounded capacity and a preference for diversity, integration still results. In this sense, the models are consistent and complementary.

Unlike “learning” models in the literature, our model does not explicitly treat information spreading perfectly from person to person. Simulations have shown that results presented here are qualitatively similar if either information decays with time or attenuates with distance (as in Zipf 1946) or is “sticky” (as in von Hippel 1998) in terms of the expertise required to process it. Either factor can move equilibrium knowledge profiles from homogeneity toward clustering, contingent on preferences. If perfect knowledge transfers are allowed, but extreme preferences prevent intergroup interaction, then subsequent results are unchanged. If learning is allowed, but balkanization refers to group formation apart from what members know, then results are also unchanged.

A contrasting perspective appears in Watts and Strogatz (1998), which models small-world phenomena. Their model considers paths between agents in which groups exhibit a high degree of local clustering but also a fairly short average path length between individuals. Through simulation and analysis, they show that adding random links to a structured network, which has high local clustering and long average path lengths, can reduce average path length much more rapidly than it reduces clustering. Thus, local communities could appear to have numerous in-group ties, while the distance to members of out-groups appears fairly short–an idea first captured in Milgram’s phrase “six degrees of separation,” implying that any two people across the globe could be linked by a chain of only six people. (1)

To the extent that data diffuses more rapidly, shorter paths between distant people will promote more integrated information. Transfer also depends, however, on preferences. Intermediate people in a chain must be willing to serve as conduits for data that need not necessarily pertain to them. In a dramatic demonstration of this, Dodds et al. (2003) tried to recreate the Milgram letter-passing experiment. Despite the ease of using e-mail over standard mail, fully 98% of chains failed to complete (Dodds et al. 2003). (2) Thus, news of popular interest, terrorist attacks, and jokes-of-the-day diffuse rapidly, while subtle ideas or those of parochial interest, like new mathematical theorems, diffuse slowly. Subtle ideas may also require sophisticated knowledge to convey. Subtle information is less likely to diffuse rapidly without loss from node to node, as the child’s game of “telephone” illustrates even for simple rumors. Related critical mass and threshold models of diffusion also appear in Valente (1995). One difference is that Valente allows for “opinion leaders,” whereas the present research treats the agents equally in the analysis.

Information integration also differs from group integration. Although the former measures the knowledge individuals have in common, the latter measures the communities they commonly form. The first considers the overlap in what people know, while the second considers the overlap in how they spend their time. As IT can affect both, we introduce measures of knowledge profiles and community membership that…

NOTE: All illustrations and photos have been removed from this article.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s