<![CDATA[Artificial intelligence (AI) deployment in critical domains requires machine learning systems that are not only accurate but also robust, interpretable, fair, and aligned with responsible governance principles. However, conventional machine learning evaluation approaches often prioritize predictive performance and computational efficiency while giving limited attention to ethical accountability, transparency, regulatory compliance, and sustainability. This study aims to develop a trustworthy machine learning evaluation framework for robust and interpretable AI systems. The focus of the study is the evaluation of intelligent systems across healthcare, finance, and transportation, where reliability and accountability are essential for real-world deployment. A qualitative case study approach was employed through expert interviews, literature analysis, document review, and cross-domain case comparisons to identify key evaluation dimensions. The findings show that trustworthy evaluation should integrate technical indicators, including accuracy, robustness, and interpretability, with broader dimensions such as fairness, accountability, governance compliance, and social responsibility. The proposed framework provides a structured model for assessing intelligent systems beyond conventional performance metrics. It also supports better consistency in interpretability assessment, stronger fairness evaluation, and improved alignment with international AI governance expectations. This study contributes to the development of responsible AI by offering a practi- cal evaluation framework that can guide researchers, developers, and institutions in designing machine learning systems that are reliable, transparent, and socially accountable. The framework has implications for sustainable and compliant AI implementation in high-impact sectors.]]>
