Pilot Study: Epsilon-Greedy Algorithm in AI-Based Web Application for Mastering Multiplication Tables

The integration of technology into education is becoming a common practice. Artificial intelligence systems, which have recently become an indispensable part of everyday life, are a natural step towards innovation in education. We have developed a web application for learning and mastering multiplication tables based on the idea of an intelligent tutoring system. The app personalizes the learning process, as it selects tasks based on an analysis of the previous user's answers. The algorithm for task selection is based on the idea of The Multi-Armed Bandit Problem which is one of the concepts in machine learning, more precisely in reinforcement learning. The main challenge in this problem is addressing the exploration-exploitation dilemma, which looks for a balance between exploring new possibilities and exploiting already known optimal solutions. There exist different strategies for addressing this dilemma. In this study, we implemented the Epsilon-greedy policy strategy. The article describes the results of a pilot study implemented on a small group of university students. We observed the algorithm's behavior regarding the diversity of the generated tasks and checked for any excessive repetition of tasks from the same number.