(Un)Fair Machine

A three-part series covering varying conceptions of fairness in the context of algorithmic decision-making. Algorithm designers may control what the system has access to (procedural fairness) or recalibrate decision criteria after a score has been produced (outcome fairness).

Each chapter opens with a visualization of relevant real-world data and asks the reader to tweak model parameters based on what they think is fair. Knotty, abstract concepts are then introduced, formalized, and examined.