Arguably, the more successful application of link prediction techniques is in the recommendation of users to follow on online social networks. This is widely witnessed by huge investments for the development of accurate and scalable user-to-user recommendation systems. However, all existing techniques fall short in providing explanations and, if you share our same experience, you may often wonder why the system gives you a recommendation for following a given user.

In this work we study link prediction with explanations for user recommendation systems. Our model not only predicts links among members of the network, but for each link it decides whether it is “topical” or “social,” and depending on this decision it provides the user with a different type of personalized explanation.