The aim of this thesis is to construct a model for evaluating training programs and to check its relevance by confronting it with empirical data. We conducted this intervention-research in the framework of a call for proposals and evaluated a social experiment bearing the characteristics of a training program. Within this context of commissioned research, we explain the emergence of an area for the research and articulation of a social dimension (for praxeological purposes) and an academic dimension (for heuristic purposes). The evaluation model is based on the method of accountability but also on the development of the actors involved. It mobilizes two different methodological orientations: “acquiring knowledge to objectify” and “acquiring knowledge to interpret”. The first takes a quantitative approach and measures the efficiency of the program; the second completes the first and facilitates its interpretation. The second orientation takes a qualitative approach measuring the cognitive developments of the training beneficiaries, characterizing the unexpected effects and efficiency as perceived by the actors involved. Thanks to the data analysis, we have been able to consolidate certain elements of the model, highlight some of its limits and offer some leverage points to help stabilize it. We propose adding a component to the second orientation “acquiring knowledge to interpret”, in order to focus on the analysis of change in the implementation of the program. Likewise, we provide five methodological, axiological, heuristic, epistemological and political leverage points which strengthen the general coherence of the model. The evaluation model has withstood the confrontation with empirical data.