Agnieszka will present a new approach to detect Bias in data. Learn about GEBI – Global Explanations for Bias Identification. The proposed method aims to detect bias in data with attribution-based locally-summarized global explanations, coming from post-hoc Explainable Artificial Intelligence (XAI).

• Is GEBI a model agnostic post-hoc approach?
• What is the value of isomap dimensionality reduction in your approach?
• What are the advantages of a global explanation approach vs. a local explanation?

Global Explanations for Bias Identification (GEBI) is counterfactual approach for detecting bias in image classifiers