We report on our development of an educational social machine based on the concept that feedback in communities is an effective means to support the development of communities of learning and practice. Key challenges faced by this work are how best to support educational and social interactions, how to deliver personalised tuition, and how to enable effective feedback, all in a way which is potentially scalable to thousands of users. A case study is described involving one to one and group music lessons in an on-campus, face to face, higher education context that were observed and analysed in terms of the actions carried out by the participants. The actions are described and it is shown how they can be formalised into a flowchart which represents the social interactions and activities within a lesson. Through this analysis, specific scenarios emerged where the feedback being given might not be effective, e.g the recipient not understanding the feedback or the provision of feedback which is not specific enough. In answer to these scenarios of ineffective feedback, the requirements for a technological intervention which aims to make the feedback more effective are proposed. With this in mind, we are then able to describe a novel technological platform which has been developed as part of a large-scale European research project and which aims to support effective feedback. The platform is based around focused discussion of time based media, embedded within existing teaching activities at a research led higher education institution in the UK. We outline how it is being used in a blended learning model to support the teaching and learning of music. We reflect on the experience of developing techniques and systems for enabling communities of e-learning and describe our evaluation methodology which involves several case studies and approximately 400 users in its current phase.