Online learning technologies are supporting the widespread access to quality education, which is a driver of economic growth and social cohesion. However, a lot of students has felt the social isolation that comes within an online learning environment, and new ways to combat loneliness are needed. Online communities can be a promising tool to favor the social engagement, but there is no approach capable of (i) creating a bond between learners by automatically composing tailored online communities, and (ii) recommending them courses that can be virtually experienced together in accordance with their needs and characteristics (e.g., knowledge level, topic interest, preferred learning style, personality similarity, future ambitions). Given the relevance of creating technologies that favor the interaction of the users on the Web to reduce online loneliness, and the importance of favoring a collaborative learning experience, we propose SociaLearning, an intelligent service that groups similar learners, and produces shared course recommendations to them. Thanks to the grouping phase, online communities that can interact are created, while by leveraging the shared course recommendations tailored to each group, each online community can follow the same course(s) jointly. Our proposal can (i) engage learners in a shared experience, going beyond classic approaches, (ii) favor the social interaction, reducing the social loneliness that typically characterizes the online experience, (iii) increase the effectiveness of the learning process of each learner, thanks to the collaborative attitude that the proposal favors, and (iv) promote marketing strategies to reduce the cost for accessing the courses, thanks to special discounts for groups, as an example. In parallel, the instructors can (i) put in practice a wide range of learning strategies based on group tasks and (ii) get a better teaching experience. Finally, the approach can affect by extension educational institutes, content providers, technology providers, and the other entities involved in the loop.