Fisher Information Principle

A measure of how sensitively a probability model's likelihood responds to changes in a parameter; formally, for a parametric model the Fisher information equals the expected value (under the model) of the squared score (derivative of the log-likelihood).