Explaining Simulations Through Self Explaining Agents

Several strategies are used to explain emergent interaction patterns in agent-based simulations. A distinction can be made between simulations in which the agents just behave in a reactive way, and simulations involving agents with also pro-active (goal-directed) behavior. Pro-active behavior is more variable and harder to predict than reactive behavior, and therefore it might be harder to explain. However, the approach presented in this paper tries to make advantage of the agents' pro-activeness by using it to explain their behavior. The aggregation of the agents' explanations form a basis for explaining the simulation as a whole. In this paper, an agent model that is able to generate (pro-active) behavior and explanations about that behavior is introduced, and the implementation of the model is discussed. Examples show how the link between behavior generation and explanation in the model can contribute to the explanation of a simulation.