Algorithmic systems are increasingly embedded in our societies – and it’s no longer good enough for reporters to declare AI models impossibly complex and unknowable. Instead, experts point out that "people make AI," and that the harms they create can be traced to corporate choices and human values. In this advanced workshop, a Pulitzer Prize-winning data journalist shows how investigating datasets that are used to train AI systems can penetrate the algorithmic black box, and expose the causes of the gaps, errors, and biases that can emerge.