Trust in automation is an essential characteristic of supervised control. Indeed, counter performance of man-machine duo caused by poor trust level (overtrust or undertrust) can lead to material or human damage. It is therefore necessary to assess the trust of the man to anticipate such situations and prevent their realization. Therefore, in this thesis, we propose an evaluation model of trust. For this, we defined two models. The first is a model of dialogue based on the theory of grounding and specific for supervised control. The second is an evaluation model of trust which describes the relationship between user trust and observation of his strategy dialogue obtained trough the first model. The theoretical contribution is accompanied by two experiments. They are contextualized in the field of multi-UAV control and more specifically for patrol and interception missions. The first experiment, with exploratory purposes, highlights the existence of a link between trust and dialogue. The second checks our evaluation model of trust. This thesis therefore provides the basis for an objective assessment of trust with the observation of man-machine dialogue.
