Abstract:

Game-based learning like serious games is growing more and more, but to be effective, these games have to be personalized according to the learning progress while keeping their playful aspect. Learning analytics techniques are usually used to collect data about a significant portion of learner activities and analyze learning progress. However, the current works on learning analytics do not include playful aspects. In another side, video games analytics techniques are restricted to entertainment to keep the player connected to the game as long as possible. So, the paper aims to propose to personalize the content to be learned in an educational and playful manner for players-learners within a same model. Three software agents compose this model, they interact with the learner through the game interface and use several data structures. The most important data is of course domain data, for instance a set of exercices that could be used by the game and the learning paths. Learner data concerns all the activities that are relevant to analyze performance, like success/failure, response time, etc., as well as psychological profile of player. Some pedagogical rules are also stored to validate learning progression, specially success conditions and importance level of a given content. The first step of the personalization consists of selecting the better game mechanics to be used for each learner, such as social function or scores table. These game mechanics can be used within any game phase. They are added to the game or else the game evolves whilst respecting player preferences. Messages will need to be predefined to interact according to these mechanics. This is the task of the telemetry agent, it also collects learner activities to analyze their performance. The personalization agent evaluates the better content to be proposed to learner according the results obtained by the telemetry agent. Finally, the visualization agent offers a help to all the users – learner, teacher, or parent – of the serious game, it can show the impact of the performed work with a graph and makes predictions on the remaining work. The model was validated by creating a prototype that verified its functionality on a breakout game to learn French grammar rules. Thus, with this model, the game can offer the most relevant content for learner, show their progress and interact according to the game mechanics that best suits each learner. Although playful personalization is limited, the model is flexible enough to adapt to any form of educational content and any field of study. Tests involving learners will be made more later and would allow a more advanced validation.

Keywords:

Serious game, learning analytics, personalization.