Interesting. I wonder if someone has tried to combine the two. I guess modern deep reinforcement learning is one such combination because it combines feedback (reinforcement) and probabilistic descriptions but maybe there are other interesting combinations of probability, causality, and feedback.