AN AGENT-BASED MODEL FOR PERSONALIZED LEARNING IN NETLOGO USING Q-LEARNING

Abstract

This study investigates the effectiveness of using the Q-learning reinforcement learning algorithm in the context of personalized student learning by constructing an agent-based model in the NetLogo simulation environment. The proposed model assumes the existence of a population of student agents, each possessing individual characteristics, especially a distinct learning style (visual, auditory, or kinesthetic). The learning environment is modeled as a grid of patches, each containing educational content of a particular type (textual, video, or audio) and a specified difficulty level.Agents interact with the environment by making decisions to either stay on their current location or move to another patch, based on an ε-greedy action selection strategy. The Q-learning algorithm enables each agent to form and continuously update its own Q-table, which stores the utility estimates of various actions in different states. The reward is determined by the degree of alignment between the agent’s learning style and the content type. If the style matches the content format (e.g., auditory style and audio material), the agent receives a positive reward; otherwise, a negative one.The results of a series of simulations showed that agents gradually learn to avoid mismatched content and begin to prefer patches that correspond to their perception style. An increase in average reward, a growing proportion of agents with positive outcomes, and a decrease in the frequency of random movements over time were observed. This indicates the formation of a stable learning policy and effective adaptation to the environmental conditions. Moreover, the influence of Q-learning parameters (ε, α, γ) on learning success was analyzed, allowing for the identification of optimal configurations for achieving the best performance. The obtained results confirm the relevance of using reinforcement learning approaches to model personalized educational strategies. The proposed model can be used as a foundation for the development of intelligent learning support systems capable of adapting the educational process to individual learners’ needs and traits.