Résumé
Role-playing games are decision-making support tools and are used to help “issues wording” by stakeholders in resources management situations. Unfortunately, the results obtained after having run games sessions are ambiguous, as stationary replications of games sessions are difficult to implement. The control of the context influence on outcomes is a good approach to assess games performances and to provide rules that can be followed when building other games. After having simplified a role-playing game (applied in a watershed in South Africa) this study shows how a game context could be decomposed and re-composed according to its main dimensions: illustration of the instructions, communication, periods repetition, and players experience, the impact of different levels of these dimensions being experimentally tested (at LEEM, “Experimental Economics Lab in Montpellier”). We assess the impact of the two first dimensions: “illustration” and “communication”, by comparing the outcomes obtained after having varied their levels with the ones obtained with the referential treatment. We show that addition of contextual elements in instructions produces a noise in behaviours, as much important as context is developed in details. However, “water management” context does not impact the players (students) decisions in the same way than other context chosen in another field (in our protocol: salaries in a firm). Finally, argumentation of their choices by players improve learning outcomes, decisions convergence until referential equilibrium is faster than without communication between players.