Stimulating collaborative activity in online social learning environments with Markov decision processes

Our work is motivated by a belief that social learning, where a community of students interact with eachother to co-create and share knowledge, is key to our students developing 21st century skills. However, convincing students to engage in and value this kind of activity is challenging. In this paper, we employ a technique from AI research called a Markov Decision Process (MDP) to model social learning activity then to suggest interventions that might increase the activity. We describe the model and its validation in simulation and draw conclusions about the effectiveness of this approach in general. The main contributions of the paper is to (i) show how it is possible to model education data as an MDP (ii) show that the resulting decision policy succeeds in guiding the system towards goal states in simulation.