Reducing Decision Uncertainty in AI-Based Student Career Guidance Using a Hybrid Machine Learning and Large Language Model Framework

Educational decision-making, particularly the selection of academic streams and career pathways, involves high levels of uncertainty and long-term consequences for students. Although machine learning–based guidance systems have demonstrated strong predictive performance, many students remain hesitant or unconvinced by algorithmic recommendations due to limited interpretability and contextual understanding. This paper presents a hybrid Machine Learning–Large Language Model (ML–LLM) framework designed to reduce decision uncertainty rather than focusing solely on prediction accuracy. The proposed system integrates supervised machine learning models for academic stream prediction, psychometric assessment, and dropout-risk analysis with an LLM-based advisory module that provides natural-language explanations and confidence-aware guidance. To evaluate system effectiveness, uncertainty-oriented metrics such as Prediction Entropy, Decision Stability Score, and Risk Reduction Index are employed alongside traditional performance measures. Experimental results based on real student data demonstrate that the inclusion of LLM-driven explanations significantly improves decision confidence and stability compared to ML-only systems. The findings highlight the importance of uncertainty-aware evaluation in educational AI systems and support the role of explanation-driven hybrid frameworks in improving student-centered decision support.