System identification appears increasingly in bio- logical system modeling. In biology, repeatability is an impor- tant factor because each system, such as an organ, provides different responses according to individuals. Therefore, it is necessary to handle this variability to infer results on the entire population. Integrating inter-individual repeatability of responses within the modeling step is possible through the introduction of mixed effets. However, no similar approach exists in the field of dynamical system identification. In this article we propose an identification method of dynamical systems with random effects represented by an ARX structure using the EM algorithm. Simulations show the interest of this approach comparing to a more classical procedure consisting to identify parameters of each person independently.