Human-AI coevolution supposes that knowledge of human and AI system evolves in the course of their activities. This knowledge evolution can be done by applying human-centered shared autonomy between human and AI systems. It includes the discovery and the updating of knowledge. To do so, this paper proposes a new feature for AI-based system: the educable learning process. It consists in making system educable by discovering and updating technical or human knowledge when inconsistency between knowledge is occurring in the course of use experience. Knowledge is modelled with natural human reasoning principles and with merged or cumulative machine learning processes. The detection of inconsistencies between knowledge generates two processing: knowledge discovery and knowledge updating. The former aims to link inputs with outputs of knowledge and to explore new possible links. The latter consists in updating existing knowledge by translating and transferring it via allocation and communication system to humans or AI systems. Educable learning process is illustrated by a practical use-case study in transportation domain by discovering and updating user’s knowledge to a cumulative AI system.