(Un)Fair Machine

A three-part series covering the different conceptions of fairness applicable to decision-making algorithms. Algorithm designers can control what the system has access to (promoting procedural fairness) or recalibrate decision criteria post-prediction (promoting outcome fairness).

Each part opens with visualizations of pertinent real-world data, asking the reader to tweak model parameters based on their own conception of fairness. The result is an engaging experience where knotty concepts are broken down and thoroughly examined.