АLGORITHMS FOR DETECTION AND PREVENTION OF MANIPULATIONS IN GAMIFIED EDUCATIONAL ENVIRONMENTS

Abstract

Gamification of educational platforms, incorporating points, achievements, and leaderboards, has proven effective in enhancing student motivation and engagement. However, the focus on extrinsic rewards introduces vulnerabilities, notably the phenomenon of farming, where users obtain rewards without genuine knowledge acquisition. This article proposes an algorithmic approach to detect and prevent farming in gamified educational environments.Objective. research and development of algorithms for detecting and preventing pharming in gamified educational environments using artificial intelligence methods. The tasks include formalizing the system, analyzing user behavior, and creating adaptive correction mechanisms.Methodology. The research methodology relies on constructing a graph-based model of user activity within a gamified educational system. Anomalous patterns are identified using the DBSCAN clustering algorithm and Local Outlier Factor (LOF) for anomaly detection. Adaptive behavioral correction is implemented through Q-learning, enabling the preservation of user motivation without intrusive interventions. The algorithm is periodically retrained based on new data, adapting to changes in user behavior.The solution architecture integrates activity graph construction, DBSCAN clustering, LOF-based anomaly detection, and adaptive reward system correction via reinforcement learning (Q-learning). The proposed approach achieves up to 95 % detection of anomalous patterns with a minimal false-positive rate. Particular emphasis is placed on soft correction–a flexible adaptation mechanism that maintains student motivation during evaluation interventions.Scientific Novelty. The novelty of the research lies in the adaptation of cybersecurity methods to the context of educational analytics, the development of algorithms for detecting and preventing pharming for gamified educational applications using artificial intelligence with the subsequent use of neural networks (RNN, transformers) to predict manipulations and implement personalized correction strategies based on user profiles.Conclusion. The proposed solution represents a significant step toward creating more ethical, transparent, and reliable gamified learning systems. The findings can be applied to develop next-generation gamified platforms that combine adaptability, resilience to manipulation, and high-quality collection and interpretation of educational data.