Emergent interactions in conditioned processes: a route to smart matter?

We propose a method to derive `smart' active-matter interactions, such as collision avoidance, from first principles. Begin from a system with naive interactions, for example hardcore exclusion, and identify the natural fluctuations which by serendipity yield a high value of some relevant constituent performance, like mobility: in the effective dynamics describing these fluctuations smart interactions emerge. As a proof of concept, we examine two one-dimensional models of active matter: the totally asymmetric exclusion process and run-and-tumble particles. We show that the energy efficiency of motion, limited by particle collisions, can be increased up to a point of diminishing returns by conditioning on high mobility. Particular to the second model is the emergence of alignment interactions that allow substantial gains in efficiency, by preventing jamming without increasing energy consumption. We close by presenting features of our approach that go beyond the two models considered.