<![CDATA[The evaluation of resume-based answers demands complex semantic understanding and knowledge structuring, making manual assessment prone to inter-rater inconsistency, while deep learning models such as BERT generally operate as a black box without adequate transparency. This study aims to design a Transparent AI-Driven Assessment Framework that integrates a BERT/RoBERTa model for resume assessment, an Explainable AI module based on SHAP and LIME, and a dual-validation mechanism based on multiple-choice analytics to verify learners' conceptual understanding. The study employs the Research and Development (R&D) method, covering needs analysis, architectural design, and the development of a prototype based on a microservices architecture interoperable with the Learning Management System. A formative evaluation of the prototype conducted on 15 respondents (7 lecturers and 8 students) yielded an average System Usability Scale score of 88.17 (Excellent category) and high perceived usefulness and transparency scores (4.78 and 4.72 out of 5), which preliminarily support the feasibility of the design, although these results remain indicative given the limited number of respondents.]]>
Copyrights © 2026
