Agentic AI for Sports Violation Detection by using Explainable AI

The critical and growing need for smart, transparent and real-time detection of ethical offenses in professional sports like match-fixing, doping and cyber misconduct is met by Agentic AI for Sports Violation detection by Utilizing Explainable AI. The proposed system introduces a new multi-agent architecture where every expert agent employs modality-specific information to detect a specific type of violation. These include a Cyber Agent employing optimized RoBERTa for text analysis and a Bio Agent and Match-Fixing Agent powered by GridSearch-optimized XGBoost. The multiple-intent user queries are processed by a central Task Specifier based on LLM which dynamically allocates tasks to the concerned agents. For ensuring transparency and interpretability for predictions the framework is designed with Explainable AI methods, namely SHAP (Shapley Additive Explanations) for structured data and token attribution based on attention for text inputs. Conceiving a single agentic framework for multi-domain sports infraction detection, including explainability to ensure stakeholder trust and performance comparison with classical models are some of the aims of the study. The experiments on synthetic data demonstrate good noisy and missing query tolerance, low inference latency and high accuracy. As compared to traditional methods, the new approach excels at explaining outcomes, working in real time, and making predictions even more faster and more accurately as well as efficient. The outcome proves that the application of this system towards automated, explainable and reliable monitoring of sports infractions is possible and scalable and will immensely aid fair play and moral policing within the sporting universe.