Design and Implementation of Reinforcement Learning-Based System for Personalized Educational Content Delivery in Mobile Applications: A Child-to-Child Educational Approach

Abstract

Due to the fast growing nature of educational technologies, increased flexibility of learning processes has been realized with characteristics such as flexibility in assessment and delivery styles to match the students’ pace on different learning styles. Progress and engagement are very crucial since they show how flexible the learning system is to adapt to the new technologies. Current conventional educational systems pose a significant challenge in implementing differentiated instruction in conjunction with the idea of group learning, especially in heterogeneous classes, making the mixed ability classroom nearly marginal. In response to these challenges, this research uses the reinforcement learning approach for personalized educational content delivery in mobile applications. The proposed system integrates the Double Deep Q-Network (DDQN) algorithm with a Child-to-Child Education Approach effectively, optimizing the learning experiences of both the individual learner and the group learner through the dynamic changes of their learning needs and communications. The system deals with important trade-offs in reinforcement learning, such as the exploitation and exploration trade-off and the balance between learner engagement and knowledge utilization. The elements of the developed methodology are a creation of an individualized content sharing system along with a peer-learning motivator introduced for increased cooperation. To evaluate the system’s reliability and performance, experiments were conducted with student groups from different government schools in Iraq: Hashem Al-Asadi School, Janat School, and Muhaila School. The proposed DDQN-based model was compared with conventional single and control groups, providing a strong validation scenario. The result shows enhanced learning achievement points, peer learning incentives, and general participation rates and affirms the adaptable application of the DDQN algorithm in education. Furthermore, it presents recommendations to educators to learn about the students who require the attention of a tutor; this, in essence, makes classroom management possible. From the educational technology domain perspective, this research offers an initial scalable, adaptive, and data-driven model for personalized and collaborative learning. The study provides a sound framework to integrate reinforcement learning for mobile and blended learning, offering ways forward to meet current issues in distributing resources and managing student engagement and offering potential research directions for future learning environments for remote and group learning.