Distributed agent-mediated knowledge management (AMKM) has attracted much attention as a way to improve knowledge sharing across the world. In AMKM, many systems can potentially interact with each other and share their knowledge while keeping their own ontology such as health care systems that handle problems of distributed experience and search engines that search distributed information in the Internet. The main problem for those systems is to make the agents understand each other adequately. Concept learning is an enabling technique. However, the semantic heterogeneity problem may occur. That is, those concepts may have been defined differently in separate ontologies and conflicts become inevitable. In order to overcome the problem of semantic heterogeneity, we present a mechanism for concept learning based on social networking that can be used to effectively resolve possible conflicts that may occur during the learning process.