Conference Agenda

Adolescent substance use is shaped by social relationships, but most research has focused on friendships and romantic relationships, overlooking the role of pre-dating romantic interest. Adolescents may adjust their substance use behaviors in response to crushes, friends, and extended social circles, such as friends of friends, potentially adopting behaviors modeled by these peers.

This study examines whether the substance use of crushes, friends and friends of friends influences adolescents’ own behaviors. Using two waves of social network data from the Peers and Emergence of Adolescent Romance (PEAR) study (2,159 Dutch high school students), we apply Stochastic Actor- Oriented Models (SAOMs) to distinguish peer influence from selection effects. Influence is modeled by the average substance use of social connections, and we test whether higher levels of alcohol, tobacco, and marijuana use among peers are associated with increases in adolescents’ own use. Preliminary results from one school show that both friends and friends of friends influence drinking, while friends of friends also influence smoking. Additionally, we will explore age differences in peer selection and influence, and test whether these mechanisms play a greater role in early or later years of secondary education. By integrating both direct and indirect social ties, this study extends research on peer influence and provides new insights into early risk factors for adolescent substance use.

We examine care networks altered by heroin in the western Tanzanian border cities of Mbeya and Mwanza to see how drug-using mutual aid groups and their relations to non-users differ or are similar in the two settings. Tanzanian heroin users’ mutual aid groups are a phenomenon we are unaware of in the West. We are interested in how drug users, alone and in their mutual aid groups, negotiate connectivity, obligation, and relations of dependence and belonging. Our social network analysis investigates trust, care and sharing in personal networks of care as the intersection of those affected by heroin through distribution, consumption, caring, recovery, and relapse experiences. We examine the connections surrounding methadone clinics and Sober Houses within the socioeconomic and political context of Tanzania. The cross-sectional personal network survey captures the people in the participant’s life, those people’s characteristics, and those people’s relationships to one another. In each city, our sample includes 60 people actively using, and 60 in recovery through methadone, sober houses, or other mechanisms (total ~240). In our examination of caring, trust, and sharing we compare of behaviors, perspectives, network characteristics, outcomes and perspectives between, for example, recovery vs. current user, new vs. long-time user, or different categories of alters. To tease out shared meaning and both transitory and stable activities, we join the nodes of a domain (e.g., mutual aid group) into a single node to see how that domain or type of actor relates to the other domains of caring practices in the network.

Background:

Exposure to e-cigarette-related content on TikTok is associated with e-cigarette and cannabis use among U.S. adolescents. However, its influence on peer tie formation remains unknown. This study examined whether exposure to e-cigarette and cannabis posts on TikTok influences tie formation among adolescents using social network analysis.

Methods:

We analyzed data from a longitudinal semi-annual survey across five waves (2022–2024) of California adolescents (Mage = 15 at baseline; ~50% female, ~50% Hispanic; N ~ 4,000 per wave) who completed in-classroom surveys and nominated up to five friends per wave. We applied RSiena models to analyze the influence of exposure to e-cigarette or cannabis posts on TikTok among egos who were never-users of e-cigarettes or cannabis at prior waves on tie formation with alters depending on their e-cigarette or cannabis use status at subsequent waves. Network, ego-level covariate (TikTok exposure), and homophily effects were assessed.

Results:

Networks were sparse with a significant tendency for reciprocal ties. E-cigarette non-users exposed to e-cigarette posts on TikTok were more likely to form ties with e-cigarette users at one time point but the effect was not consistent across waves. TikTok-exposed non-users of e-cigarettes/cannabis were significantly more likely to form ties with other similarly exposed non-users, indicating a consistent homophily effect.

Conclusions:

Findings suggest that adolescent e-cigarette/cannabis non-users exposed to e-cigarette/cannabis TikTok posts cluster together, but exposure does not consistently promote ties with e-cigarette or cannabis users. Network dynamics should be monitored, especially since TikTok exposure is associated with adolescent substance use.

Substance use behaviours, alcohol consumption and many other health behaviours often occur simultaneously, suggesting some latent variable which affects the manifestation of different patterns of co-occurrence. Latent class analysis and similar mixture modelling techniques allow researchers to identify underlying subgroups with similar patterns of behaviour from a broader population. These have become increasingly popular approaches in studies of poly-substance use, anti-social behaviours, adolescent delinquency, and other potentially harmful patterns of co-occurring health behaviours.

However, is also well understood that patterns of health behaviour are shaped by the social contexts of individuals and subject to processes of social influence. Currently, little is known about potential biases arising from estimating latent classes without accounting for social network dependencies. There also has been a growing interest in exploring how behaviours may spread as ‘bundles’, with patterns of co-occurring practices diffusing across networks. Integrating latent class analysis into social network modelling offers a promising approach to this area of research, but further research is needed to assess the feasibility of combining these distinct frameworks.

We present findings from ongoing work exploring how latent class analysis and related models can be integrated with social network analysis. Using empirical data on adolescent health behaviours and school friendships as well as simulation-based approaches, we assess the reliability of latent class estimation in the presence of peer network effects. We also demonstrate methods to account for classification error when incorporating latent classes as nodal attributes in social network models.

This study examines the role of peer-perceived substance use in shaping adolescent friendship and popularity networks, addressing gaps in prior research that rely on self-reported substance use or aggregated peer perceptions. Using longitudinal social network data from the SNARE project, a study of Dutch adolescents, we apply Stochastic Actor-Oriented Models (SAOMs) to analyze how classmates' perceptions of alcohol and tobacco use influence their social ties within classrooms.

Findings indicate that peer-perceived alcohol use significantly predicts friendship nominations, aligning with peer clustering theory, which suggests that shared behaviors reinforce group norms. In contrast, peer-perceived smoking does not consistently influence friendships, highlighting potential differences in how substance use behaviors are socially interpreted. For popularity networks, both alcohol and tobacco use are positively associated with nominations, supporting social identity theory and normative influence theory, which emphasize the social appeal of risk-taking behaviors. These findings suggest that adolescents perceived as substance users may be socially rewarded through increased visibility and status among peers.

By integrating a peer perception approach with a network analysis framework, this study contributes to understanding how substance use is embedded within adolescent social structures. The results underscore the importance of addressing peer perceptions in interventions targeting adolescent substance use, as well as the social mechanisms that sustain these behaviors. Understanding these dynamics can inform policies and prevention programs to reduce substance use appeal and promote healthier peer relationships.