911³Ô¹Ï

Synopsis

Collective dynamics in active matter are often described in terms of prescribed local interaction rules. An alternative perspective is to view such systems as adaptive, where interaction rules are not fixed but emerge through learning. Here, we explore this idea experimentally using active colloidal systems in which reinforcement principles shape collective behavior. We show that learning in many-particle systems naturally gives rise to emergent dynamics governed by non-reciprocal interactions and memory. In our systems, these features originate from fluid-mediated coupling and delayed environmental responses, which encode information about past states and break effective action–reaction symmetry. As a result, the collective dynamics become history-dependent and adaptive, extending beyond descriptions based on instantaneous local rules. At the same time, we demonstrate that such systems can be harnessed for computation: the intrinsic memory of hydrodynamically coupled active colloids enables the prediction of complex, including chaotic, temporal signals and the detection of subtle changes in their dynamics. These results highlight active matter as a physical platform in which learning, non-reciprocity, and memory jointly give rise to emergent behavior, providing a bridge between nonequilibrium statistical physics and concepts from artificial intelligence.