The latest trend of incorporating various data-centric machine learning (ML) models in software-intensive systems has posed new challenges in the quality assurance practice of software engineering, especially in a high-risk environment. ML experts are now focusing on explaining ML models to assure the safe behavior of ML-based systems. However, not enough attention has been paid to explain the inherent uncertainty of the training data. The current practice of ML-based system engineering lacks transparency in the systematic fitness assessment process of the training data before engaging in the rigorous ML model training. We propose the use of Dempster Shafer theory and its modified combination rule (Yager’s rule) to assess the collective confidence in the quality of a training dataset. With the example of training datasets for pedestrian detection of autonomous vehicles, we demonstrate how the proposed approach can be used by the stakeholders with diverse expertise to combine their beliefs in the quality arguments and evidences about the data. Our results also open up a great scope of future research in the field of data requirements engineering that can facilitate evidence-based data assurance for ML-based safety-critical systems.