Recently, there has been a lot of research on decentralized social networks motivated by privacy concerns with centralized systems. However, an open challenge with decentralized systems is the actual data replication and retrieval among the participating devices. We argue this aspect can be greatly improved in terms of efficiency by taking social relationships, user behavior and locality into account. User-centric networking is a paradigm which includes the users' own devices - from smartphones to highly available personal clouds - and targets at a socio-aware data storage, based on the users' behavior and their devices' availability. In this paper, we introduce the concept of a Decision Engine that chooses the replication devices based on multi-dimensional input parameters, from momentary conditions to long-time learned user behavior and social relationships.