Discovering Maximum Entropy Knowledge

Abstract

This paper seeks to determine a rational agent’s evidential constraints given her beliefs. Rationality is here construed as adherence to a principle of entropy maximisation. I determine the rational agent’s set of probability efunctions compatible with the evidence, $${\mathbb {E}}$$ E , given the maximum entropy function and given some constraints on the shape of $${\mathbb {E}}$$ E . I also consider agents employing a centre of mass approach to form their beliefs rather than entropy maximisation.