Homophily and Contagion Are Generically Confounded in Observational Social Network Studies

“Or rather are they close in the network because of these similarities, through the processes known assortative mixing on traits, or more simply as homophily (McPherson et al., 2001)?”

“…inter-connected clusters of nodes, with minimal connection between clusters — has been thoroughly explored in the recent literature (explicitly in Girvan and Newman (2002); Newman and Girvan (2003); Bickel and Chen (2009); Porter et al. (2009); Fortunato (2010), implicitly in much smaller…”

“…of nodes, with minimal connection between clusters — has been thoroughly explored in the recent literature (explicitly in Girvan and Newman (2002); Newman and Girvan (2003); Bickel and Chen (2009); Porter et al. (2009); Fortunato (2010), implicitly in much smaller clusters in Elwert and…”

“The “easy” solution is to use expert knowledge of the system to identify all causally relevant variables, measure a sufficient set of them, and adjust for them appropriately (Morgan and Winship, 2007; Pearl, 2000; Spirtes et al., 2001).”

“…with minimal connection between clusters — has been thoroughly explored in the recent literature (explicitly in Girvan and Newman (2002); Newman and Girvan (2003); Bickel and Chen (2009); Porter et al. (2009); Fortunato (2010), implicitly in much smaller clusters in Elwert and Christakis (2008)).”

“…of actual processes of transmission and influence, which have been extensively studied in social psychology and allied fields since the 1920s (Bartlett, 1932; Sperber, 1996; Huckfeldt et al., 2004; Friedkin, 1998).13 However, not only is it adequate to demonstrate the existence of the…”

“19”

“…confounded by factors such as homophily—the tendency of individuals to associate with those who share similar characteristics—raising the question of whether prevalent moods in networks result from social transmission or self-selection (Aral, Muchnik, & Sundararajan, 2009; Shalizi & Thomas, 2011).”

“Because the Hawkes process only measures temporal clustering, some events may be attributable to homophily or common shocks rather than contagion [32].”

“Recognizing that the effect we observe could instead be the consequence of homophily and linguistic accommodation due to social network position (Shalizi and Thomas 2011; Danescu-Niculescu-Mizil, Gamon, and Dumais 2011), we conclude our analysis by investigating whether our results can be explained…”

“Without further assumptions that restrict the spillovers in some way, many average causal quantities of interest cannot be identified in general [29, 27].”

“As in observational network studies more generally, peer correlation is not separable from contagion without additional assumptions (Shalizi and Thomas 2011).”

“This pattern parallels longstanding debates in sociology concerning influence versus selection [1, 34].”

“…so covariate effects are not misattributed to network structure, using permutation-or simulation-based inference, and probing robustness to omitted variables via placebo tests or negative controls (e.g., Aral, Muchnik, and Sundararajan 2009; Shalizi and Thomas 2011; Uppala and Desmarais 2023).”

“Moreover, when units within the same cluster share unobserved characteristics that also affect outcomes, cluster-level randomization may fail to balance these characteristics between treatment groups, thereby obscuring the true treat-ment effect (Shalizi and Thomas, 2011; Ugander and Yin, 2023).”

“It has been shown that such incidental similarity often leads to misinterpretation, and single-view models are prone to treat it as authentic [21].”

“Third, in diffusion and influence studies, disentangling homophily from social influence remains methodologically intractable: do users connect because they share similar views, or do they adopt similar views through connection [23]?”

“…: Neighboring regions exhibit similar infection trends due to shared demographics, mobility structures, and By bridging the gap between epidemic forecasting and policy evaluation, our approach ensures that policymakers obtain unbiased estimates of intervention effectiveness, improving public…”

“This phenomenon is often referred to as network dependence or homophily (Shalizi and Thomas 2011; Basse and Airoldi 2018; Zhang and Kang 2022).”

“Latent partial times thus incorporate distributional changes due to peer effects, but are usually correlated because characteristics influence both the choice of peers and the distribution of times, inducing, e.g. , homophily bias (Shalizi and Thomas, 2011).”

“…fundamentally different mechanisms of social behavior: influence suggests that individuals change their attitudes or behaviors in response to their network connections, while homophily suggests that individuals with similar predispositions are more likely to connect in the first place 14 .”

“While in longitudinal data the results of such tests may be taken as evidence regarding the causal effect of social influence, such interpretation has been cautioned against when applying to cross-sectional data (29).”

“Note that one cannot use these data to infer causality, due to the (latent) homophily (21).”

“Moreover, distinguishing between homophily and the influence of peers becomes challenging, especially when an unobserved factor contributes to both processes (Shalizi and Thomas 2011; Ma, Krishnan, and Montgomery 2015).”

“Our findings indicate that latent homophily threatens the inference for interference effects, corrob-orating previous theoretical and empirical results established from toy models and computer-generated network data [35, 25, 26].”

“This similarity can occur through a variety of forces, and (especially with cross-sectional data), it can be very difficult to determine which is driving the autocorrelation (Shalizi and Thomas 2011).”

“The two phenomenon are generally confounded with each other in observational data (Shalizi and Thomas, 2011).”

“Thus, to accurately identify causal peer influence, the effect of peers’ behaviors should be distinguished from the effect of similarities, which may induce similar behaviors among peers [Shalizi and Thomas, 2011].”

“Social influence theory has been used to explore the temporal relationships between behaviour of individuals at a given time with interactions in a neighbourhood network leading to suggesting direct influence of neighbour on behavioural outcome (Shalizi and Thomas, 2011).”

“Homophily may occur on any number of traits, observed or latent, and is known to confound problems of causal inference in the social sciences [38; 36; 11; 23].”

“Although longitudinal data on interaction networks are increasingly available (2, 16), separating network formation from peer influence using observational data has been statistically challenging due to various endogeneity concerns (17, 18).”

“Such clever incorporation appear also elsewhere, namely [20], [21], [6][4], [22], [2], [3].”

“1 The entanglement between these different mechanisms unavoidably induces bias when we estimate the influence effect (Shalizi & Thomas, 2011).”

“Their key contribution is to formalize intuition about the challenges that network interference poses for causal inference, even in experimental settings, much as prior work has done for observational network studies (Shalizi and Thomas 2011).”

“The characteristics in social networks not only define the behaviors of these entities but also depict the structure of entities that are connected to each other through some relations, for example, common interest and similar behaviors [2]–[4].”

“presented a paper on sociological methods which states homophily and contagion are generically confounded in observational social network studies [9].”

“Stepping up and down the ladder is achieved by repeatedly hitting • > to lower the threshold and obtain greater protection • to raise the threshold and lose protection.”

“Despite the sophistication of these model specifications on handling observed homophily, Shalizi and Thomas (2011) argued that peer influence and homophily are generically confounded observationally and can not be readily disentangled from each other, because latent homophily may still remain as a…”

“However, the validity of analyses based on observational studies has been called into question by several authors [10] [11].”

“An issue that has been troubling those studying online communities and social networks is distinguishing the effects of homophily, social influence, and external common causes [7], [8], [9].”

“55 behaviour of their peers) or homophily (individuals affiliating with those who act in a similar way) or a mixture of both (84).”

“Previous work suggested that independent decision making (specifically homophily) and influence can be indistinguishable in certain circumstances [23, 134], but the disagreement with our null model may provide a way to distinguish these two mechanisms.”

“Contagion/influence causes a strong friendship between two persons to pass attributes of one person to another person (induction) [9, 46].”