Sédrick Stassin, a PhD student from UMONS, led a collaborative research project with Géraldin Nanfack (UNamur), Nassim Versbraegen (ULB), and several other researchers at the TRAIL Summer Workshop 2022 in Berlin, Germany 🇩🇪.

The project, titled "Bias Detection on Image or Text Data and their Mitigation on Models," was presented at the Belgian Embassy in Berlin. This work is part of the Grand Challenge 6: “Trusted AI Systems” of TRAIL, focusing on developing AI models with enhanced transparency, fairness, and non-discrimination. 🤖✨

The research team proposed a comprehensive framework for detecting biases in image and text data and mitigating them when training machine learning models. Unlike previous studies that assume prior knowledge of bias sources, this approach targets unstructured data, such as images and text, where biases are often hidden and unknown.

The project integrates advanced bias detection and mitigation techniques with post-hoc explainability methods. equips deep learning practitioners with tools to detect and handle biases using popular datasets like Colored MNIST and biased text datasets, providing essential solutions for critical applications in fields like healthcare, finance, and beyond.