This paper explores the cybernetic regulation of complex human learning and teaching. It provides a theoretical description of the arrangement of adaptive, machine- generated learner feedback which relies on cybernetic principles. Cybernetics — today often referred to as control theory, or feedback control theory — involves the incorporation of self-establishing feedback mechanisms for optimal control in complex systems. Although feedback is considered a key element of any learning process, the arrangement of feedback by teachers and educators is under pressure because of the ever-growing complexity of learning environments which is being reinforced by open, online learning technologies and topical models of learning (competence learning, experiential learning, situated cognition, serious gaming). This paper explores how cybernetic principles could be implemented in complex learning environments, e.g. serious games, for the arrange- ment of self-regulating feedback loops for learners. The approach is based on a quanti- tative description of learning activities and learning performances. For the presentation of the feedback, a feedback decision procedure is suggested which is to be linked with pedagogical theories and assessment models. The proposed cybernetic approach is elucidated with a theoretical example. The paper provides a proof of principle and gives suggestions for further development.