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”

“, [2]), a structural analysis of the causal graphical model [38] encoding influence and homophily shows that distinguishing them is impossible [42].”

“Social influence is a separate, but related concept explaining how individuals encourage similarity through their interactions [14].”

“In some sense, homophily could thus potentially “explain away” triadic closure, much in the way homophily is confounded with social influence (i.e. peer effects) more generally ( 41 ).”

“This distribution can be a product of social and spatial differentiation, as people have different preferences and resources.”

“Therefore, it is complicated to establish causality (Elwert and Winship 2014; Mouw 2006; Shalizi and Thomas 2011).”

“Further, true random samples from the population of interest are not always easy to obtain; in fact, the process of gathering data is often far from ideal, as it may involve unknown sampling biases [10, 11, 59] or network effects [12, 60].”

“Hence, if friendship selection depended on some unobserved covariate that also predicted smoking initiation, our estimates would be biased because we would not be able to disentangle selection from influence (Shalizi and Thomas 2011).”

“The social contagion (the effects caused by the people close to each other in a social network) are shown to be confounded by the effect of an individual’s covariates on his or her behavior or other measurable responses (Shalizi & Thomas, 2011).”

“…a network (Girvan & Newman, 2002; Holland, Blackmond, & Leinhardt, 1983), finding structures in network connections (Frank & Strauss, 1986; Robins, Pattison, Kalish, & Lusher, 2007), and understanding how the network connections influence behaviors of units in the network (Shalizi & Thomas, 2011).”

“Moreover, homophily in unobserved attributes (latent homophily) that may be indicated by the very existence of a relationship of peers with a common friend (see Shalizi and Thomas 2011) will be equally represented in recipient and non-recipient peer groups.”

“Under special conditions and with panel data this disentanglement is possible (see Bisin and Özgür, 2012; or Shalizi and Thomas, 2011 for a negative result in the most general case).4 We contribute to this literature bringing evidence on the importance of the elicitation of friendship network to…”

“…works include causal diagrams for interference (Ogburn & VanderWeele, 2014), structural equation models (SEMs) (Ogburn et al., 2022), chain graphs (Shpitser et al., 2017; Ogburn et al., 2020), structural causal models (SCMs) at the unit level with linear structural equations (Zhang et al.,…”

“Establishing causal effects with observational data is challenging, as it is difficult to eliminate all non-causal explanations (Shalizi and Thomas 2011).”

“This means that the process of social influence, the confounding by shared exogenous social environments of social networkmembers, and endogenous social selection cannot be isolated from each other (Shalizi and Thomas, 2011).”

“2013; Kossinets and Watts 2009; Shalizi and Thomas 2011).”

“Although participants could potentially state that they learned from an influential or important person, revealed knowledge sharing is supported by similarity in medicinal plant knowledge [43, 49].”

“For instance, Shalizi and Thomas (2011) argued that controlling for alter’s lagged obesity status may at best only partially account for homophily bias.”

“Therefore, in future communication on the app they would likely have a shared starting point for connection and conversation and have a sense of familiarity which, in turn, could become a driver for future communication (Shalizi and Thomas, 2011).”

“Recently, some researchers report the social theories related to topic categories in online social media, such as homophily (McPherson, Smith-Lovin, and Cook 2001), consistency (Abelson 1983) and contagion (Shalizi and Thomas 2011; Harrigan, Achananuparp, and Lim 2012).”

“Moreover, at the methodological level the social network ‘contagion’ hypothesis has also been critically discussed [12,21,22,23].”

“Graph datasets may demonstrate homophily and heterophily properties [16–19].”

“If agents are homophilous in μi, then this captures unobserved homophily, a well-known hindrance to identifying social interactions (Shalizi and Thomas, 2011).”

“, thus, it might be impossible to verify whether ties similarity is really the result of homophily or social influence [24].”

“Hence distinguishing between the effects of social influence and homophily at the individual level remains a challenge [34, 35].”

“For more discussion on this issue, for example, see [2, 94, 76, 97].”

“…peer influence has been a focus of research for decades (e.g., Kandel, 1978), a number of methodological issues (Lyons, 2011; Noel & Nyhan, 2011; Shalizi & Thomas, 2011) have plagued traditional and contemporary work on this topic; and it is only relatively recently that methodological advances…”

“contemporaneous peer effects (Lyons 2011) and the identification of peer effects independent of homophily (Shalizi and Thomas 2011).”

“Shalizi and Thomas (ST) illustrate how the confounding effects of latent homophily cause standard tests for contagion to fail with two examples [14].”

“Yet, it is difficult to disentangle homophily from social influence [5, 6].”

“…similar to themselves as friends, socialization, wherein individuals become more similar to their friends over time, or shared external causation (Shalizi and Thomas, 2011) Research in an early adolescent sample (ages 11 – 13) suggests that these results may also extend beyond domain-general…”

“Firstly, researchers have been aware of the potential bias of results due to social homophily (Shalizi & Thomas, 2011).”

“Addressing collider bias in the causal modelling of social network data and learning-based hypotheses presents additional challenges (Lyons, 2011; Shalizi & Thomas, 2011).”

“Note that these two theories are not mutually exclusive nor conficting, although quantifying social contagion is challenging when it is usually unavoidably contaminated by network homophily [65].”

“This process could unfold via either social homophily or social contagion or both (Shalizi and Thomas 2011).”

“For instance, a study by Shalizi et al. on social network data has found that the use of regression coefficients, even taking into consideration temporal data, cannot distinguish causal factors from confounding variables [18].”

“Observational social network studies have not gone without criticism, though, mainly due to them having difficulty separating causal social influence from other effects such as homophily (Shalizi and Thomas 2011) and context effects (Cohen-Cole and Fletcher 2008).”

“Shalizi and Thomas (2011) point out challenges arising from the confounding of social contagion, homophily, and the influence of individual traits.”

“While friend similarities are often interpreted as evidence of peer effects, causality is hard to demonstrate because of alternative pathways to behavioural similarity, such as shared ecology and similarity-based assortment, that are often difficult to tell apart on the basis of observational data [76].”

“This line of work is drawing inspiration by social correlation theories such as homophily and social influence [1], [2].”

“All too often, however, observational data alone fail to distinguish not only between social learning and individual learning, but also between social learning and homophily—the tendency for people with similar traits to co-associate [19,20].”

“However (Shalizi and Thomas 2011) have shown that homophily and social influence are generically confounded in social diffusion processes and it in general not easy to distinguish between the two effects.”

“…as with earlier work on peer effects in economics, causal inference for effects of online social content is difficult due to endogeneity in which users see signals from whom (Shalizi and Thomas 2011) as well as which users decide to generate social content (Toubia and Stephen 2013, Taylor et al. 2013).”

“When global patterns emerge, their origin is often contested (Watts and Strogatz; 1998; Shalizi and Thomas; 2011).”

“…of initial, pre-exposure agreement that remains stable over time, for example, all group members expressing happiness instantaneously due to shared social identities and group-based appraisals or homophily and similar response tendencies (see Figure 1; Mackie & Smith, 2015; Shalizi & Thomas, 2011).”

“In fact, some authors have argued that in observational studies, contagion effects cannot be reliably differentiated from the effects of homophily, which refers to the tendency of individuals to select friends who resemble them (Shalizi & Thomas, 2011).”

“Scarce, but important network studies have refined the argument to show how certain types of homogeneous relationships, such as work relationships, are especially conducive for fostering other types of relationships, like friendships (Liberman et al., 2021; Shalizi & Thomas, 2011).”

“The theory of secondary homophily holds that [29] adolescents tend to befriend peers with the same ethnic and racial backgrounds [30], similar school grades [31,32], and similar socioeconomic status [33] and these attributes potentially intersect with educational expectations, and might increase the expectation-relate…”

“With crosssectional data, it is impossible to differentiate between selection and influence (Shalizi and Thomas 2010).”

“Real social networks present three potential confounders to social influence [49, 50].”

“Unfortunately, disentangling selection and social influence have proven to be extremely difficult (Shalizi and Thomas 2011; VanderWeele 2011).”

“A group of friends protesting a mutual acquaintance’s decision to buy a fuel-inefficient SUV can be seen as 1 See Christakis and Fowler (2013) and Shalizi and Thomas (2011) for a recent exploration of the identification problem in the econometric analysis of diffusive processes in social networks,…”

“, thus, it might be impossible to verify whether ties similarity is really the result of homophily or social influence [24].”

“Several commonly recognized influence factors in social sciences include homophily, contagion, and causality [27], [29] where homophily captures the similarities among social”

“Dividing graphs into homogeneous, densely inter-connected clusters of nodes, with minimal connection between clusters has been the topic of community discovery [31], [35], [39][42].”

“In the case of public health policy, for example, positive evidence for the role of social contagion in the spreading of such disparate attributes as happiness [20], obesity [21], and loneliness [22], have been challenged due to their reliance on under-sampled reconstructed social networks [23].”

“…is an extensive literature on causal inference methods for so called “network settings” from a variety of perspectives (Hudgens and Halloran, 2008; Shalizi and Thomas, 2011; Ogburn and VanderWeele, 2014; van der Laan, 2014; Sofrygin and van der Laan, 2017; DeAmour, 2016; Aronow and Samii, 2017;…”