Responsible AI in student management: preventing misdecision in career choice of university students under inaccurate guidance.

In today's digital era, university students' career-related decision-making has become increasingly complex, shaped not only by personal aspirations and societal expectations but also by the algorithmic recommendations embedded in AI-supported technologies. While such systems provide convenient references, their outputs are often vague, generalized, or contextually limited, which underscores the irreplaceable role of human judgment in situations of uncertainty. Against this backdrop, the present study develops a multi-criteria decision-making (MCDM) framework tailored to the context of student management and career guidance. Specifically, it integrates the compromise ranking strategy CRADIS (Compromise Ranking of Alternatives from Distance to Ideal Solution) with the expressive flexibility of Circular Pythagorean Fuzzy Sets (Cir-PyFS). Cir-PyFS was chosen to reflect the modalities of non-decision, inexactness, and subjective desire in human evaluation that people may have. The linguistic assessments of each alternative were redefined in CPFS representations within the proposed Circular Pythagorean Fuzzy-based CRADIS model (Cir-PyFBCM), which incorporates membership, non-membership, and radius values. The model outperformed traditional fuzzy MCDM approaches in terms of stability, interpretability, and sensitivity to the influence of decision-makers in an empirical exercise using a realistic career-planning example with several alternative courses of action, multiple expert decision-makers, and eight benefit-based evaluation criteria. These comparison and sensitivity analysis results support that the proposed methodology serves as an effective bridge, linking mathematical precision with the human approach to reasoning, and offers a more careful, flexible, and reliable tool for governing high-stakes decisions, especially university students, when making their career decisions in AI-supported realms. More broadly, the study underscores the potential of responsible AI-assisted decision-making in promoting reflective, evidence-based practices within higher education management.