We are studying disagreement in the annotation of legal data, aiming to distinguish error from genuine subjectivity and build better datasets. This builds off existing questions on methods of annotation for expert datasets (e.g. Conceptual Questions in Developing Expert Annotated Data)