I forget where I've seen this, but I like using a squiggly arrow to represent a function that returns a probability.

For example, in reinforcement learning, the policy $$\pi$$ can be a stochastic map from states to probabilities of an action.

$$\pi : S \rightsquigarrow A$$