Consider the Jeffreys Prior for a parameter $\theta$. It is defined as $\pi(\theta) \propto \sqrt{I(\theta)}$. Why is this prior often preferred in Bayesian inference?