The global prevalence of social media has led to dramatic changes in the contemporary media environment and how people engage in civic and political lives. This burgeoning trend has sparked fierce debate over the role that social media plays in the functioning of democracy. The present study aims to address this debate by drawing on two theoretical approaches: network social capital and cross-cutting discussion. Based on a theoretical distinction between these two approaches, this study is able to unravel two underlying mechanisms through which apolitical use of social media (e.g., relational use of Facebook and WhatsApp) can generate substantial political consequences at both attitudinal and behavioral levels. More importantly, these two mechanisms point to the center of the “democracy paradox,” suggesting the possibility that social media may help cultivate an informed and tolerant citizenry while impair a civic culture featured with active engagement (i.e., civic engagement and political participation).

Specifically, the first mechanism presumes positive indirect effects of relational use of Facebook on citizen engagement and generalized trust via increased network diversity as well as adverse indirect effects of relational use of WhatsApp on the same outcomes via decreased network diversity. The second mechanism posits a null effect of relational use of Facebook and two opposite indirect effects of relational use of WhatsApp on citizen engagements (negative effect) and political tolerance (positive effect) through the mediation of increased online cross-cutting discussion.

Leveraging online survey data (n = 1,322) targeting Hong Kong residents aged 18–54 years old, an empirical study was implemented to test both mechanisms. Survey administration was conducted by a reputational company, Dynata, from June to July 2020. Linear regression and causal inference based on inverse probability weighting were applied to model testing.

Results show that Facebook’s relational use fosters citizen engagements and generalized trust by diversifying one’s social network (namely the first mechanism), while WhatsApp’s relational use is found to encourage citizen engagements through increased online cross-cutting experiences (namely the second mechanism). These findings indicate that apolitical usage (namely relational use) of Facebook and WhatsApp can generate different political consequences through divergent mechanisms, which advances our knowledge of the democratic implications that social media have. Our findings also shed light on the prospect of democratic politics in Hong Kong.