The past few years have witnessed the rapid growth of online social networks, which have become important hubs of social activity and conduits of information. Identifying social influence in these newly emerging platforms can provide us with significant insights to better understand the interaction behaviors among online users. However, it is difficult for us to measure the influence quantitatively among user peers, since many key factors such as homophily and heterogeneity, can not be observed in our real world conveniently. More recent work mainly focuses on developing theoretical models based on explicit causal knowledge. Nevertheless, such knowledge is usually not available and often needs to be discovered. In this paper, we introduce a model free approach to formulate causal inferences of behaviors among user peers. Experimental results show that influence measured by our approach could successfully reconstruct the underlying networks structure. Furthermore, two additional case studies based on this approach reveal that influentials wield power through specific venues, which constitute a comparatively small portion of the whole channels.