Beyond Accuracy: Building Fairer ML Models with Python

Over the last few years, machine learning systems have increasingly been used to make decisions that affect people and their access to opportunities. As such, understanding and addressing fairness-related risks has become an essential part of building responsible AI systems. In this 45-minute lecture, participants will be introduced to the key concepts of fairness and bias in machine learning. Through practical examples, we will examine how bias can enter data, propagate through model development, and appear in model predictions. Finally, participants will have an opportunity to use Python to measure bias and build a simple mitigation model to improve fairness in machine learning systems. By the end of the session, participants will be better equipped to identify, measure, and begin addressing unfair outcomes in their own machine learning projects.