A Multimodal Framework for Explainable Evaluation of Soft Skills in Educational Environments
To address the substantial bias and poor interpretability inherent in current soft skills assessment practices in higher education, this study proposes an interpretable evaluation framework integrating multimodal perception with fuzzy logic. Methodologically, it pioneers the fusion of a Granular Linguistic Model of Phenomena (GLMP) with computer vision–driven facial expression and gesture recognition to enable fine-grained modeling and explicit quantification of uncertainty for competencies such as decision-making, communication, and creativity. The primary contributions are: (1) the first semantic-interpretability–enabled and uncertainty-transparent soft skills assessment paradigm; (2) empirical validation showing significant improvements in scoring consistency (+32.7%) and expert comprehensibility among undergraduate students; and (3) demonstrably superior assessment quality from multimodal integration versus unimodal approaches, yielding traceable, educationally actionable outputs.