User assistance for serious games using hidden markov model

Serious Games, i.e., games not just for pure entertainment and with characterizing goals, are gaining huge popularity for the purpose of education and training. To further increase the learning outcome of serious games, assistance functionalities like adaptive systems observe the users and try to guide them to achieve their learning objectives. The research question is how to model the user's behavior, their progress, and how to determine the best adaptation strategies to motivate the users and provide assistance whenever required. Using experience-data in a serious game is one approach to develop and train models for adaptivity. In this paper, we present SeGaAdapt, an adaptive framework that is based on a Hidden Markov Model (HMM) algorithm for providing dynamic user-assistance and learning analytics for a serious game. For the development and training of the HMM, we use activity streams or user interaction data gained from an Experience API (xAPI) tracker. The adaptivity mechanism uses the HMM to analyze the current state of the user (player) in order to predict the best feasible activity for future states. Technical verification of this work-in-progress implementation shows the feasibility of the approach and hints at future research directions.