RESEARCH ON INTERPRETABLE MACHINE LEARNING AND EXPLAINABLE AI MODELS FOR PREDICTING APPLICANTS CAREER DECISION-MAKING

Abstract

The integration of artificial intelligence into human resource management and vocational guidance has catalyzed a paradigm shift from traditional, counselor-driven career advisement to highly complex, data-driven predictive modeling. While sophisticated machine learning architectures—ranging from ensemble decision trees to deep neural networks—demonstrate unprecedented accuracy in mapping psychometric profiles, academic records, and labor market intelligence to optimal career trajectories, their inherent structural opacity poses significant ethical, regulatory, and pedagogical challenges. The "black-box" nature of these predictive engines obscures the fundamental logic behind career recommendations and hiring rankings, thereby risking the exacerbation of historical biases, eroding user trust, and violating emerging international regulatory frameworks. This comprehensive analysis evaluates the theoretical foundations, methodological deployment, and practical implications of Explainable AI (XAI) frameworks, specifically focusing on SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME), as critical mechanisms for achieving interpretable machine learning in career decision-making. By systematizing the intersection of established behavioral psychology paradigms—such as Holland’s RIASEC model, the Five-Factor Model of personality, and Social Cognitive Career Theory (SCCT)—with advanced algorithmic interpretability, this report demonstrates how XAI transitions predictive models from opaque algorithmic gatekeepers into transparent, developmental tools. The empirical evidence synthesized within this review suggests that while interpretable machine learning enhances predictive fidelity and mitigates subgroup differences, its most profound value lies in fostering human-AI collaboration, auditing algorithmic fairness, and empowering applicants through transparent, data-driven self-efficacy. By shifting the focus from mere predictive accuracy to pedagogically meaningful explanation, organizations and educational institutions can ensure that AI-driven career guidance remains an equitable, reliable, and legally compliant instrument for global workforce development.