Understanding and Addressing Fairwashing in Machine Learning

Fairwashing is the risk where an unfair black-box model is explained by a fairer model using post-hoc explanation manipulation. These attacks can generalize and transfer across different black-box models, even without explicitly using their predictions, complicating detection. The presentation reviews current research on mitigating the potential for fairwashing. The speaker is Sébastien Gambs, Professor of Computer Science at UQAM and Canada Research Chair in Privacy-preserving and Ethical Analysis of Big Data, specializing in privacy, fairness, and accountability in machine learning.