Herding Cats: Assessing Complex Procedural and Tacit Knowledge with Multi-agent Simulations

Complex procedural knowledge and tacit knowledge are hard to assess and practice constructs. In this paper, we demonstrate the application of autonomous multi-agent simulations to assess teacher’s classroom management skills. We developed a visual virtual environment within which virtual agents mimic students. These agents act and interact autonomously with each other. Human teachers are requested to interfere on the basis of a pre-defined set of actions to optimize the simulated study time and happiness of the agents. This paper presents analyses of systematically configured simulation runs to elucidate the simulation characteristics and to evaluate the rationality and credibility of the simulations.