The MIROR Project deals with the development of an innovative adaptive system for music learning and teaching based on the reflexive interaction paradigm. The platform will be developed in the context of early childhood music education. It will act as an advanced cognitive tutor, designed to promote specific cognitive abilities in the field of music improvisation, both in formal learning contexts (kindergartens, primary schools, music schools) and informal ones (at home, kinder centres, etc.). The reflexive interaction paradigm is based on the idea of letting users manipulate virtual copies of themselves, through specifically designed machine-learning software referred to as interactive reflexive musical systems (IRMS). By definition IRMS are able to learn and configure themselves according to their understanding of learner’s behaviour. We propose to extend the IRMS paradigm with the analysis and synthesis of multisensory expressive gesture to increase its impact on the musical pedagogy of young children, by developing new multimodal interface. The project will be based on a novel spiral design approach involving coupled interactions between the technical partners and the psycho-pedagogical ones. The project will integrate both psychological case-study experiments, aiming to experiment cognitive hypothesis concerning the mirroring behaviour and the learning efficacy of the platform, and validation studies aiming at developing the software in concrete educational settings. The project will contribute to promoting the reflexive interaction paradigm not only in the field of music learning but more generally as a new paradigm for establishing a synergy between learning and cognition in the context of child/machine interaction. The project mostly addresses target c) concerning the development of adaptive and innovative learning systems. However, objectives of target d) are also pursued through the development of a novel music learning appliance.