Simulated Learners for Testing Agile Teaming in Social Educational Games

This paper proposes an approach for creating and testing an multiagent systems based adaptive social educational game (SEG), QuizMASter, using the concept of simulated learners to overcome experimentation complexity and unpredictable student availability, as is typical with online learning environments. We show that simulated learners can play two roles. First, it can be used for testing the game planning, scheduling and adaptive assessment algorithms. With some degree of success met with our initial experimentation with QuizMASter, advanced planning and coordination algorithms are now needed to allow the gamebased assessment platform to realize its full potential. The multi-agent system approach is suitable for modeling and developing adaptive behaviour in SEGs. However, as we have found with our early prototypes, verifying and validating such a system is very difficult in an online context where students are not always available. MAS-based assessment game planning and coordination algorithms are complex and thus need simulated learners for testing purposes. Second, to overcome unpredictable student availability, we modeled QuizMASter as a new class of sociotechnical system, human-agent collective (HAC). In the system, human learners and simulated learners (smart software agents) engage in flexible relationship in order to achieve both their individual and collective goals, while simulated learners are selected for serving as virtual team members.