Social relationships are important among persons experiencing homelessness, but there is little research on changes in social networks among persons moving into permanent supportive housing (PSH). Using data collected as part of a longitudinal study of 405 adults (aged 39+) moving into PSH, this study describes network upheaval during this critical time of transition. Interviews conducted prior to and after three months of living in PSH assessed individual-level (demographics, homelessness history, health, and mental health) and social network characteristics, including network size and composition (demographics, relationship type, and social support). Interviewers utilized network member characteristics to assess whether network members were new or sustained between baseline and three months post-housing. Multilevel logistic regression models assessed characteristics of network members associated with being newly gained or persisting in networks three months after PSH move-in. Results show only one-third of social networks were retained during the transition to PSH, and veterans, African Americans, and other racial/ethnic minorities, and those living in scattered site housing, were more likely to experience network disruption. Relatives, romantic partners, and service providers were most likely to be retained after move-in. Some network change was moderated by tie strength, including the retention of street-met persons. Implications are discussed.

Understanding how person-to-person contagious processes spread through a population requires accurate information on connections between population members. However, such connectivity data, when collected via interview, is often incomplete due to partial recall, respondent fatigue, or study design, e.g. fixed choice designs (FCD) truncate out-degree by limiting the number of contacts each respondent can report. Research has shown how FCD affects network properties, but its implications for predicted speed and size of spreading processes remain largely unexplored. To study the impact of degree truncation on predictions of spreading process outcomes, we generated collections of synthetic networks containing specific properties (degree distribution, degree-assortativity, clustering), and used empirical social network data from 75 villages in Karnataka, India. We simulated FCD using various truncation thresholds and ran a susceptible-infectious-recovered (SIR) process on each network. We found that spreading processes on truncated networks resulted in slower and smaller epidemics, with a sudden decrease in prediction accuracy at a level of truncation that varied by network type. Our results have implications beyond FCD to truncation due to any limited sampling from a larger network. We conclude that knowledge of network structure is important for understanding the accuracy of predictions of process spread on degree truncated networks.

A person’s egonet, the set of others with whom that person is connected, is a personal sample of society which especially influences that person’s experience and perceptions of society. We show that egonets systematically misrepresent the general population because each person is included in as many egonets as that person has “friends.” Previous research has recognized that this unequal weighting in egonets leads many people to find that their friends have more friends than they themselves have. This paper builds upon that research to show that people’s egonets provide them with systematically biased samples of the population more generally. We discuss how this ubiquitous egonet bias may have far reaching implications for people’s experiences and perceptions of frequencies of other people’s ties and traits in ways that may influence their own feelings and behaviors. In particular, these egonet biases may help explain people’s tendencies to disproportionately experience and overestimate the prevalence of certain types of deviance and other social behaviors and consequently be influenced toward them. We illustrate egonet bias with analyses of all friends among 63,731 Facebook users. We call for further empirical investigation of egonet biases and their consequences for individuals and society.

We investigate the impact of misinformation about the contact structure on the ability to predict disease outbreaks. We base our study on 31 empirical temporal networks and tune the frequencies in errors in the node identities or time stamps of contacts. We find that for both these spreading scenarios, the maximal misprediction of both the outbreak size and time to extinction follows an stretched exponential convergence as a function of the error frequency. We furthermore determine the temporal-network structural factors influencing the parameters of this convergence.

Graph kernels for learning and inference on sparse graphs have been widely studied. However, the problem of designing robust kernel functions that can effectively compare graph neighborhoods in the presence of noisy and complex data remains less explored. Here we propose a novel graph-based kernel method referred to as an edit distance graphlet kernel. The method was designed to add flexibility in capturing similarities between local graph neighborhoods as a means of probabilistically annotating vertices in sparse and labeled graphs. We report experiments on nine real-life data sets from molecular biology and social sciences and provide evidence that the new kernels perform favorably compared to established approaches. However, when both performance accuracy and run time are considered, we suggest that edit distance kernels are best suited for inference on graphs derived from protein structures. Finally, we demonstrate that the new approach facilitates simple and principled ways of integrating domain knowledge into classification and point out that our methodology extends beyond classification; e.g. to applications such as kernel-based clustering of graphs or approximate motif finding. Availability: www.sourceforge.net/projects/graphletkernels/

For sexually transmitted infections like HIV to propagate through a population, there must be a path linking susceptible cases to currently infectious cases. The existence of such paths depends in part on the degree distribution. Here, we use simulation methods to examine how two features of the degree distribution affect network connectivity: Mean degree captures a volume dimension, while the skewness of the upper tail captures a shape dimension. We find a clear interaction between shape and volume: When mean degree is low, connectivity is greater for long-tailed distributions, but at higher mean degree, connectivity is greater in short-tailed distributions. The phase transition to a giant component and giant bicomponent emerges as a positive function of volume, but it rises more sharply and ultimately reaches more people in short-tail distributions than in long-tail distributions. These findings suggest that any interventions should be attuned to how practices affect both the volume and shape of the degree distribution, noting potential unanticipated effects. For example, policies that primarily affect high-volume nodes may not be effective if they simply redistribute volume among lower degree actors, which appears to exacerbate underlying network connectivity.

Metric graph properties lie in the heart of the analysis of complex networks, while in this paper we study their convexity through mathematical definition of a convex subgraph. A subgraph is convex if every geodesic path between the nodes of the subgraph lies entirely within the subgraph. According to our perception of convexity, convex network is such in which every connected subset of nodes induces a convex subgraph. We show that convexity is an inherent property of many networks that is not present in a random graph. Most convex are spatial infrastructure networks and social collaboration graphs due to their tree-like or clique-like structure, whereas the food web is the only network studied that is truly non-convex. Core–periphery networks are regionally convex as they can be divided into a non-convex core surrounded by a convex periphery. Random graphs, however, are only locally convex meaning that any connected subgraph of size smaller than the average geodesic distance between the nodes is almost certainly convex. We present different measures of network convexity and discuss its applications in the study of networks.

One of the fundamental principles driving diversity or homogeneity in domains such as cultural differentiation, political affiliation, and product adoption is the tension between two forces: influence (the tendency of people to become similar to others they interact with) and selection (the tendency to be affected most by the behavior of others who are already similar). Influence tends to promote homogeneity within a society, while selection frequently causes fragmentation. When both forces act simultaneously, it becomes an interesting question to analyze which societal outcomes should be expected.

To study this issue more formally, we analyze a natural stylized model built upon active lines of work in political opinion formation, cultural diversity, and language evolution. We assume that the population is partitioned into “types” according to some traits (such as language spoken or political affiliation). While all types of people interact with one another, only people with sufficiently similar types can possibly influence one another. The “similarity” is captured by a graph on types in which individuals of the same or adjacent types can influence one another. We achieve an essentially complete characterization of (stable) equilibrium outcomes and prove convergence from all starting states. We also consider generalizations of this model.

The paper explores the relationship between personality characteristics, sense of community and the structure of personal networks. We collected personal network data from 100 adults, consisting of the information about their 45 alters and ties among them. In addition to the typically used bivariate correlations to analyze the relationship between individual psychological differences and network structure, we propose first the use of typologies of networks and personality and second modified versions of the triadic census in ego-networks. The results show that different personality types tend to occupy a different kind of networks and that newly introduced triadic measures show a relatively higher association with examined psychological attributes than global network measures. Overall, the Psychological Sense of Community showed higher associations with network measures than Big Five personality traits. The former was positively correlated with measures of closed triads among alters, but also with alters embedded in triads with weak ties. Regarding personality traits, Emotional Stability was positively correlated with strong closed triads and with the overall indicator of density. The elaboration of typologies and the census of triads are shown to be effective strategies for the description of personal networks, as well as for the analysis of individual psychological differences.

We review the empirical comparison of Stochastic Actor-oriented Models (SAOMs) and Temporal Exponential Random Graph Models (TERGMs) by Leifeld & Cranmer in this journal [Network Science 7(1):20–51, 2019]. When specifying their TERGM, they use exogenous nodal attributes calculated from the outcome networks’ observed degrees instead of endogenous ERGM equivalents of structural effects as used in the SAOM. This turns the modeled endogeneity into circularity and obtained results are tautological. In consequence, their out-of-sample predictions using TERGMs are based on out-of-sample information and thereby predict the future using observations from the future. Thus, their analysis rests on erroneous model specifications that invalidate the article’s conclusions. Finally, beyond these specific points, we argue that their evaluation metric—tie-level predictive accuracy—is unsuited for the task of comparing model performance.

The dynamic drivers of interfirm interactions across space have rarely been explored in the context of disaster recovery; therefore, the mechanism through which shocks propagate is unclear. This paper uses stochastic actor-oriented modeling to examine how trade networks among the 500 largest Japanese companies evolved during 2010 and 2011, i.e. before and after the Great East Japan Earthquake to identify sources of vulnerability in the system. In contrast to previous reports on broken supply chains, the network displayed only modest change even in the directly affected areas. Controlling for distance and for firm size, we find that when firms changed their partners, they preferred firms that were popular among other firms, that had partners in common with them and that also bought some products or services from them. These findings concur with a criticism that Japanese firms avoid external actors and exhibit inflexibility in reorganizing their networks in times of need, which contrasts with the non-cliquish network structures observed in high-performing economic sectors. The results also highlight the role of energy firms in disaster resilience. Unlike other large Japanese companies that cluster in major urban centers, energy firms are distributed across Japan. However, despite their peripheral physical locations, energy firms are centrally located in trade networks. Thus, while a disaster in any region may affect some energy firms and lead to large-scale temporary shocks, the entire network is unlikely to be disconnected by any region-specific disaster because of the spatial distribution of the topological network core formed by energy companies.

Research in the area of social networks and health has demonstrated that lay social network members play a critical role in the early stages of the illness career, influencing key decisions and pathways to formal care. Here, we revisit and extend this body of work, examining how the lay social network context can moderate the influence of treatment experiences on recovery outcomes as the illness career unfolds. To achieve this goal, we address two research questions, drawing on a longitudinal sample of people making their initial contact with the mental health treatment system: First, we explore how treatment experiences, lay social network characteristics, and recovery outcomes change over 2 years, beginning with the point of entry into treatment. Second, we examine whether the relationship between perceived treatment experiences and recovery outcomes is contingent on characteristics of the lay network context in which clients are socially embedded, focusing on the network's cultural orientation toward medical professionals. We find that positive treatment interactions facilitate improved self-esteem, mastery, role functioning, recovery optimism, and global functioning when the lay network culture is pro-medical, but largely have null effects on the recovery process when the lay network is more hostile to medical professionals.

Network scholars commonly encounter multiple networks, each of which is possibly governed by distinct generation rules while sharing a node group structure. Although the stochastic blockmodeling—detecting such latent group structures with group-specific connection profiles—has been a major topic of recent research, the focus has been given to the assortative group discovery of a single network. Despite its universality, concepts, and techniques for simultaneous characterization of node traits of multilayer networks, constructed by stacking multiple networks into layers, have been limited. Here, we propose a Bayesian multilayer stochastic blockmodeling framework that uncovers layer-common node traits and factors associated with layer-specific network generating functions. Without assuming a priori layer-specific generation rules, our fully Bayesian treatment allows probabilistic inference of latent traits. We extend the approach to detect changes in block structures embedded in temporal layers of network time series. We demonstrate the method using synthetic multilayer networks with assortative, disassortative, core-periphery, and overlapping community structures. Finally, we apply the method to empirical social network datasets, and find that it detects significant latent traits and structural changepoints. In particular, we uncover endogenous historical regimes associated with distinct constellations of states in United States Senate roll call vote similarity patterns.

Though declining since the 1990s, adolescent pregnancy remains common in the United States. Social supports appear to improve outcomes for pregnant teens; however, teen pregnancy introduces social obstacles, such as stigma. This study investigates how currently or previously pregnant teens’ friendship networks differ from nonpregnant girls using data from the National Longitudinal Study of Adolescent to Adult Health (Add Health) and multilevel regression models. To mitigate concerns that background differences contribute to both pregnancy risk and social networks, girls who experience a pregnancy prior to one data collection time point are compared girls who experience their first pregnancy after this time point. This group who become pregnant after the time point is presumably more similar to already pregnant teens than those never experiencing teen pregnancy. When compared to these girls who become pregnant in the future, those who have already experienced a teen pregnancy report similar numbers of friends (out-nominations) and perceived social acceptance, but are predicted to have fewer peers reporting them as friends (in-nominations) and fewer reciprocated friendships. This suggests that pregnant teens may face stigmatization, of which they may be unaware. It further highlights a new comparison group to account for selection in studies of adolescent pregnancy.

Antisocial behavior can be contagious, spreading from individual to individual and rippling through social networks. Moreover, it can spread not only through third-party influence from observation, just like innovations or individual behavior do, but also through direct experience, via “pay-it-forward” retaliation. Here, we distinguish between the effects of observation and victimization for the contagion of antisocial behavior by analyzing large-scale digital trace data. We study the spread of cheating in more than a million matches of an online multiplayer first-person shooter game, in which up to 100 players compete individually or in teams against strangers. We identify event sequences in which a player who observes or is killed by a certain number of cheaters starts cheating and evaluate the extent to which these sequences would appear if we preserve the team and interaction structure but assume alternative gameplay scenarios. The results reveal that social contagion is only likely to exist for those who both observe and experience cheating, suggesting that third-party influence and “pay-it-forward” reciprocity interact positively. In addition, the effect is present only for those who both observe and experience more than once, suggesting that cheating is more likely to spread after repeated or multi-source exposure. Approaching online games as models of social systems, we use the findings to discuss strategies for targeted interventions to stem the spread of cheating and antisocial behavior more generally in online communities, schools, organizations, and sports.

Graph embedding is a transformation of nodes of a network into a set of vectors. A good embedding should capture the underlying graph topology and structure, node-to-node relationship, and other relevant information about the graph, its subgraphs, and nodes themselves. If these objectives are achieved, an embedding is a meaningful, understandable, and often compressed representation of a network. Unfortunately, selecting the best embedding is a challenging task and very often requires domain experts. In this paper, we extend the framework for evaluating graph embeddings that was recently introduced in [15]. Now, the framework assigns two scores, local and global, to each embedding that measure the quality of an evaluated embedding for tasks that require good representation of local and, respectively, global properties of the network. The best embedding, if needed, can be selected in an unsupervised way, or the framework can identify a few embeddings that are worth further investigation. The framework is flexible and scalable and can deal with undirected/directed and weighted/unweighted graphs.

Community-based physical activity programs, such as the Recreovía, are effective in promoting healthy behaviors in Latin America. To understand Recreovías’ challenges and scalability, we characterized its social network longitudinally while studying its participants’ social cohesion and interactions. First, we constructed the Main network of the program’s Facebook profile in 2013 to determine the main stakeholders and communities of participants. Second, we studied the Temporal network growth of the Facebook profiles of three Recreovía locations from 2008 to 2016. We implemented a Time Windows in Networks algorithm to determine observation periods and a scaling model of cities’ growth to measure social cohesion over time. Our results show physical activity instructors as the main stakeholders (20.84% nodes of the network). As emerging cohesion, we found: (1) incremental growth of Facebook users (43–272 nodes), friendships (55–2565 edges), clustering coefficient (0.19–0.21), and density (0.04–0.07); (2) no preferential attachment behavior; and (3) a social cohesion super-linear growth with 1.73 new friendships per joined user. Our results underscore the physical activity instructors’ influence and the emergent cohesion in innovation periods as a co-benefit of the program. This analysis associates the social and healthy behavior dimensions of a program occurring in natural environments under a systemic approach.

This paper elaborates on two theories of social influence processes to multiplex network structures. First, cohesion influence is based on mutual communication made by different types of relations, and second comparison influence that is built on contrasting types of tie. While a system of bundles with a mutual character constitutes the setting for a multiplex network exposure measure within cohesion, comparison influence is defined algebraically through classes of actors in terms of a weakly balanced semiring structure that considers positive, negative, and also ambivalent types of tie. A case study with these approaches is made on an entrepreneurial community network with formal business relations, informal friendship ties, and perceived competition among the firms, and the methods are validated with the Sampson Monastery data set.

State preferences play an important role in international politics. Unfortunately, actually observing and measuring these preferences are impossible. In general, scholars have tried to infer preferences using either UN voting or alliance behavior. The two most notable measures of state preferences that have flowed from this research area are ideal points (Bailey et al., 2017) and S-scores (Signorino & Ritter, 1999). The basis of both these models is a spatial weighting scheme that has proven useful but discounts higher-order effects that might be present in relational data structures such as UN voting and alliances. We begin by arguing that both alliances and UN voting are simply examples of the multiple layers upon which states interact with one another. To estimate a measure of state preferences, we utilize a tensor decomposition model that provides a reduced-rank approximation of the main patterns across the layers. Our new measure of preferences plausibly describes important state relations and yields important insights on the relationship between preferences, democracy, and international conflict. Additionally, we show that a model of conflict using this measure of state preferences decisively outperforms models using extant measures when it comes to predicting conflict in an out-of-sample context.