Nonconformity in Online Social Networks: Experimental Evidence

We explore the role of social influence in online social networks. In a large-scale field experiment on a Chinese social networking site, we find that after minimizing con- founding factors such as homophily, observational learning and identity signaling to out-group members, the likelihood of a user adopting a particular choice may decrease with the adoption rate of that choice among her friends. The direction of social influence depends on user characteristics: minority and new users are, for example, more likely to react positively to an increased adoption rate than the majority and well- established users. We then replicate the study in a real-world social advertising setting that allows for learning. We find that while increased adoption is indeed more likely to lead to conformity, well-established users could still decrease the likelihood to conform as adoption rate increases. The results suggest that users in online social networks carefully trade off their needs to belong and to maintain their individuality, and the efficacy of social media advertising depends critically on user characteristics. This is joint work with Monic Sun of Boston University and Feng Zhu of Harvard Business School.