We discuss auditing computational models of human behavior and the large-scale datasets they are trained on. Our guest gives examples of the consequences of a badly curated model.