VKontakte Study Reveals Friends-of-Friends Drive 2x More Likes

A new preprint from Russian researchers Ivan Kozitsin and Anton Proskurnikov (arXiv:2606.23530) dives into the mechanics of social contagion on likes using a massive dataset of ~290,000 users from VKontakte, Russia's largest social platform. The study separates direct influence (from friends) from indirect influence (from friends-of-friends) to understand how liking behavior spreads through a network. Using statistical models, the authors controlled for demographic and structural factors like gender, degree, and activity levels, and found that the strongest predictors of a user liking a post are the actions of nodes at distances 1 and 2 in their network.

Remarkably, the influence of second-order neighbors (friends-of-friends) persists even when the user has no active direct neighbors—meaning the effect is truly indirect, not just a chain of direct influences. This suggests that seeing a like from someone two steps away can be enough to trigger a like, possibly through awareness or social norms. The researchers found no significant effect beyond distance 3. They also validated the structural diversity hypothesis: likes are more likely when a user's active friends belong to multiple distinct connected components (i.e., different friend circles), rather than one big clique. These findings have practical implications for platform design, recommendation algorithms, and viral marketing strategies.