On Selection Bias and Fairness Issues in Machine Learning

Attend this seminar presentation exploring the critical intersection of selection bias and fairness in machine learning systems. Examine how biased data collection and sampling procedures can lead to discriminatory outcomes in algorithmic decision-making, with Professor Stephan Clémençon from Telecom Paris providing theoretical foundations and practical insights. Discover statistical methods for detecting and mitigating selection bias while ensuring fairness across different demographic groups. Learn about the mathematical frameworks that govern fair machine learning, including concepts of equalized odds, demographic parity, and individual fairness. Analyze real-world case studies demonstrating how selection bias manifests in various applications such as hiring algorithms, credit scoring, and healthcare prediction models. Explore advanced techniques for bias correction and fair representation learning that can be implemented in production systems. Understand the trade-offs between accuracy and fairness, and gain insights into regulatory and ethical considerations surrounding algorithmic fairness. This presentation is part of the "Representing, calibrating & leveraging prediction uncertainty from statistics to machine learning" program at the Isaac Newton Institute.