The paper develops an argument for the criteria that a theory of ignorance should meet. It starts from the distinction between instrumental and non-instrumental action. Usually, the latter is considered irrational and the former rational as being based upon known cause-effect relations whilst the latter is not. I argue that the former requires a reasoned basis in predictive knowledge of cause and effect, without which good council is either for inaction or noninstrumental action. The argument proceeds by exploiting mainstream statistical methods to explore an example of a 'metric of advised ignorance' to guide explicit reasoned choice allowing rejection of instrumental action in favour of inaction or non-instrumental action. The argument then explores a case study of how such rejection is disallowed by official requirements in International Development Assistance (aid) that contexts must always be believed predictive and so action organised as instrumental. This shows the basic irrationality of mainstream policy rationality. The paper then discusses wider social epistemological issues of this irrationality and concludes with a list of criteria a theory of ignorance should meet.