The standard theory of decision making under uncertainty advises the decision maker to form a statistical model linking outcomes to decisions & then to choose the optimal distribution of outcomes. It assumes that you trust the model completely. But what should a decision maker do if the model cannot be trusted? This work answers this question.