DIMACS/DyDAn Workshop on Network Models of Biological and Social Contagion

November 3 - 4, 2008
DIMACS/DyDAn Center, CoRE Building, Rutgers University

Lauren Ancel Meyers, University of Texas at Austin, laurenmeyers at mail.utexas.edu
Michelle Girvan, University of Maryland, girvan at umd.edu
Presented under the auspices of the Special Focus on Computational and Mathematical Epidemiology and the Center for Dynamic Data Analysis (DyDAn).


Delia Baldassarri, Princeton University

Title: Dynamics of Political Polarization

D. Baldassarri, P. Bearman " Dynamics of Political Polarization," American Sociological Review, 72:784-811.

This article accounts for two puzzling paradoxes. The first paradox is the simultaneous absence and presence of attitude polarization--the fact that global attitude polarization is relatively rare, even though pundits describe it as common. The second paradox is the simultaneous presence and absence of social polarization--the fact that while individuals experienced attitude homogeneity in their interpersonal networks, their networks are characterized by attitude heterogeneity. These paradoxes give rise to numerous scholarly arguments. By developing a formal model of interpersonal influence over attitudes in a context where individuals hold simultaneous positions on multiple issues we show why these arguments are not mutually exclusive and how they meaningfully refer to the same social setting. The results from this model provide a single parsimonious account for both paradoxes. The framework we develop may be generalized to a wider array of problems, including classic problems in collective action, social networks and diffusion.


Damon Centola, MIT

Title: New Theory and Experiments on Diffusion in Social Networks

The strength of weak ties is that they tend to be /long/ ? they connect socially distant locations. Research on "small worlds" shows that these long ties can dramatically reduce the "degrees of separation" of a social network, thereby allowing ideas and behaviors to rapidly diffuse. However, I show that the opposite can also be true. Increasing the frequency of long ties in a clustered social network can also inhibit the diffusion of collective behavior across a population. For health related behaviors that require strong social reinforcement, such as dieting, exercising, smoking cessation, or even condom use, successful diffusion may depend primarily on the /width /of bridges between otherwise distant locations, not just their length. I present formal and computational results that demonstrate these findings, and then propose an experimental design for empirically testing the effects of social network topology on the diffusion of health behavior.

Aaron Clauset, Santa Fe Institute

Title: Hierarchically modular structure in complex networks

Many studies suggest that many complex networks exhibit not only modular structure, in which vertices divide into groups, but hierarchical structure, where these groups further subdivide into groups of groups, and so forth over multiple scales. In many cases these groups correspond to known functional units, such as ecological niches in food webs, modules in biochemical networks (protein interaction networks, metabolic networks, or genetic regulatory networks), or communities in social networks.

In this talk, I'll describe a generative model of hierarchical structure, and a general technique based on maximum likelihood for inferring such hierarchical structure directly from network data. I'll also show results indicating that hierarchies can simultaneously explain and quantitatively reproduce many commonly observed topological properties of networks, such as right-skewed degree distributions, high clustering coefficients, and short path lengths. As time allows, I'll also discuss the prospect of using similar techniques to make other kinds of inferences about the large-scale organization of networks.

Petter Holme, KTH Royal Institute of Technology

Title: Models of the co-evolution of networks and information flow

Two social processes with some degree of empirical evidence are that: I. People that are acquainted influence the opinions of each other. II. The probability that one agent becomes acquainted depends on the opinion of both. Points I and II form a feedback loop that, in more general terms, couple the dynamics on the network (the opinion spreading) to the dynamics of the network itself, and back again. I will discuss a few models of such feedback processes and general questions about the modeling of social processes. Are they just non-falsifiable intellectual entertainment, or can they help us understand our society, or even predict its future?

Thomas House, University of Warwick

Title: Capturing contact structure for respiratory disease

While some infectious diseases are capable of airborne or environmental transmission, others (including influenza and smallpox) have close contact between individuals as their preferred mode of transmission. This poses questions fistly about the structure of human contact patterns and secondly about how to capture these patterns appropriately in mathematical models. Contact surveys and pairwise models are, respectively, possible answers to these questions.

Carl Latkin, Johns Hopkins School of Public Health

Title: Relationships between Social Norms, Social Network Characteristics, and HIV Risk Behaviors in Thailand and the U.S.

Social norms have been associated with a wide range of health behaviors. We examined whether the social norms of HIV risk behaviors are clustered within social networks of injection drug users and whether the norms of network members are linked to the risk behaviors of their social network members. We also assess the outcomes of an experimental intervention designed to diffused behavior change within risk networks. The experimental intervention consisted of six small group peer-educator training sessions and two booster sessions delivered to the network index only. Data were collected from 354 networks with 933 participants in the Philadelphia, US and Chiang Mai, Thailand. Four descriptive HIV risk norms of sharing needles, cookers, and cotton and front or back-loading among friends who inject were assessed. Three of four injection risk norms (sharing needle, cookers, and cotton) were found to be significantly clustered. In Philadelphia, one network member's (the index participant) norms of sharing needles and front or back-loading were found to be significantly associated with the network members' risk behaviors, and the norm of sharing cotton was marginally associated. Index members in the intervention arm engaged in more conversations about HIV risk following the intervention compared to control indexes (OR = 1.42, p = 0.004). There was no evidence of change in sexual risk as a result of the intervention. Reductions in injection risk behaviors were observed: 37%, 20%, and 26% reduction in odds of sharing cottons, rinse water and cookers respectively, and 24% reduction in using a syringe after someone else. Analysis of the individual sites suggested a pattern of reductions in injection risk behaviors in the Philadelphia site. In both sites, the intervention resulted in index IDUs engaging in the community role of discussing reduction in HIV injection risk behaviors. The results of this study suggest that among injection drug users, social norms are clustered within networks; social networks are a meaningful level of analyses for understanding how social norms lead to risk behaviors, providing important data for intervening to reduce injection related HIV risks.

David Liben-Nowell, Carleton College

Title: Tracing Information Flow Online

Although the continuous circulation of information, news, jokes, and opinions is ubiquitous in the worldwide social network, the actual mechanics of how any single piece of information spreads on a global scale have largely remained mysterious. A major challenge lies in the difficulty of acquisition of large-scale data recording the diffusion of any particular piece of information. In this talk, I will survey some recent computational research that studies information propagation through the digital traces of online activity. I will focus on recent joint work with Jon Kleinberg that traces such information-spreading processes via the reconstruction of the propagation of two recent massively circulated Internet chain letters, one protesting the beginning of the Iraq war and one protesting budget cuts for public radio.

Miller McPherson, Duke University

Title: Social Contagion in Blau Space: Imputing social context in survey data

We develop a method of imputing characteristics of the network alters of respondents in probability samples of individuals using the homophily principle to estimate the properties of core discussion networks. These properties include a measure of the potential exposure to the attitudes, values, beliefs, and other characteristics of the respondents network alters. Data from the General Social Survey data demonstrate that the imputed network characteristics are strongly related to individual level measures such as attitudes, beliefs, and other variables typical of survey analysis. In some cases, the imputed network variable drastically alters and even eliminates the effects of standard sociodemographic variables such as age and education. We argue that the imputed network variable captures many of the aspects of social context that been at the core of sociological analysis for decades.

Chandra Muller, University of Texas at Austin

Title: Using Networks to Study the Social Dynamics of Mathematics Coursetaking in High School

This study examines how high school boys' and girls' academic effort, in the form of math coursetaking, is influenced by members of their social contexts. We argue that adolescents' social contexts are defined, in part, by clusters of students (termed "local positions") who take courses that differentiate them from others. Using course transcript data from the recent Adolescent Health and Academic Achievement Study, the authors employ a new network algorithm to identify local positions in 78 high schools in the National Longitudinal Study of Adolescent Health. Incorporating the local positions into multilevel models of math coursetaking, we find that girls are highly responsive to the social norms in their local positions, which contributes to homogeneity within and heterogeneity between local positions.

Babak Pourbohloul, British Columbia Centre for Disease Control

Title: Modeling communicable disease spread: New tools for an old problem

The underlying contact structure among individuals that determines the pattern of disease transmission and the progression of this pattern over time are two crucial elements in understanding and controlling communicable disease spread within a social setting. Recent advances in mathematical modeling have provided us the means to tackle emerging respiratory infectious disease concerns head-on; these new tools can play a crucial role in public health decision-making at times of crisis. This talk aims to share the concepts and applications of these new techniques with scientists and decision-makers interested in communicable disease control. We demonstrate how these tools can be used to address pandemic influenza preparedness as well as preparedness plans for the natural or deliberate release of infectious agents during a major public event.

Jonathan Read, University of Liverpool

Title: Measuring human social networks from the perspective of disease transmission

Surprisingly little is known about the structure of human social networks that permit the transmission of close- or casual-contact infections. In this talk I shall review some of surrogate measures employed to infer transmission networks, and present some findings from a detailed study amongst a work based peer-group within the UK. One of the striking results of this study is the large number of transient encounters made. This study also suggests that the stability of links between individuals over time can be related to the social context of the encounters, and the implications for disease transmission are explored using simulation models. This and other more recent studies among similar peer groups hint at the useful extent of contact tracing as a method of disease control.

Richard Rothenberg, Georgia State University

Title: Network Structure and function: Current perplexities

In the 75 years since network diagrams first appeared on the scene, considerable theoretical and empirical work have advanced the field. Both approaches have their difficulties and detractors, perhaps because each examines a different piece of the puzzle. Empirical work? the example here is an amalgamation of 15 completed network studies of HIV and STD transmission?has established some important hypotheses about the relationship of network structure to transmission dynamics. Transmitting networks appear to share certain common (fixed) factors: (1) a degree distribution with a long tail to the right (power law distribution, possibly scale free), indicating a disproportionate number of persons with many contacts; (2) small world characteristics, with short path lengths between persons; (3) a large connected component, suggesting a high level of reachability for most persons in a given network. Other important network characteristics are variable: transitivity (or clustering), recursion, assortativity, and concurrency all vary considerably in these data. Though this constellation of findings may characterize some transmitting networks, other structures may play a role in different settings. For example, data from parts of Africa that have experienced intense epidemic spread of HIV, suggest that the power law distriubtion of contacts may not apply, but that a pattern of low-degree, high-concurrency, long-duration relationships may be dominant. Coupled with theoretical work, and emerging information on biologically influences on infectivity, this pattern may account for rapid spread. For either of these patterns, or others that may emerge, it is difficult to demonstrate empirical quantitative relationships, and it may be that our goal should involve pattern recognition rather than precise solutions.

Erik Volz, The University of California- San Diego

Title: Serosorting in Dynamic Contact Networks

Mathematical epidemiology has progressed in the direction of combining increasingly realistic models of disease progression with increasingly complex host population structure. Contact networks have become a popular model of population structure for hosts embedded in long-term relationships. Concurrent advances in mathematical theory have made these models more tractable, which decreases the need to compromise between realism and mathematical analysis. I review several special cases of epidemics spreading through complex populations which reduce to simple solutions based on ordinary differential equations. This reveals a link between traditional mass action models of epidemics in which contacts are instantaneous and uncorrelated, and networks which have dynamically rearranging ties. I then investigate several applications. A simple modification to our equations allows us to model sero-sorting of HIV positive individuals, which is the tendency to rearrange relationships to those with similar HIV status. These models are tested using data of injection drug users in Tijuana, Mexico, and a high-risk cohort from Atlanta, Georgia.

Chris Wiggins, Columbia University

Title: Modules, communities, and wannabees: inferring the structure and scale of complex networks

The past decade has witnessed an explosion of interest in revealing coarse-grained or macroscopic structure in a variety of real world networks. Primary among the structural goals has been revealing differentially-connected regions, variously called "communities", "modules," or "clusters". With this explosion of interest has arrived a proliferation of disparate algorithms, often with parameters which ultimately tune the resolution and complexity of the resulting structure. Here we re-cast this problem as a special case of an old and well-studied problem in statistics, namely inferring latent variables under a generative model, for which the determination of ``best" parameters and model complexity can be addressed under the framework of Bayesian model selection. We illustrate how the stochastic block model can be used to reveal communities of nodes typified by their association patterns as well as the number of these communities and which of two competing ``community models" is the better for a given network.

Previous: Program
Workshop Index
DIMACS Homepage
Contacting the Center
Document last modified on October 27, 2008.