Inference, Learning, and Laws of Nature

Although inference and learning arise traditionally from different schools of thought, in the last few years they have been framed in nice unified frameworks, in the attempt to resemble clever human decision mechanisms. In this paper, however, we support the position that a true understanding of human-based inference and learning mechanisms might arise more naturally when replacing the focus on logic and probabilistic reasoning with that of cognitive laws, in the spirit of most variational laws of Nature. To this end, we propose a strong analogy between learning from constraints and analytic mechanics, which suggests us that agents living in their own environment obey laws exactly like those of particles subjected to a force field.