Clinician are interested in interpreting biomedical time series data to understand patient records. This paper proposes a new co-interpretation approach based on a "structural coupling" between two autonomous agents (a clinician and a machine) to support clinicians in their interpretation task. From an Enactive point of view, the interpretation can not be autonomously performed with direct knowledge exchanges. Then, annotation seems to be a core concept to enable mutual understanding and to preserve the required autonomy of each interpreter. The CATS system has been developed to be coupled with a clinician. It autonomously annotates time series data and evolves through its annotation experience.