This document discusses selecting alternative updating policies for fault diagnosis under a cognitive theory framework. It proposes a methodology that assigns prior probabilities to a computational decision making system based on its primary and secondary decisions regarding normal and fault conditions. The system adopts a probability distribution as its credence function. Expected utility functions are used to evaluate how well the system's beliefs match the truth, and the policy that maximizes this function is selected. As an example, it shows applying an alternative primary decision making policy based on parameter distributions and re-evaluating results. Probabilities are then assigned to the system's output based on its performance.