Behavioral flexibility reshapes prosocial punishment under spatial reciprocity

Prosocial punishment is thought to thrive under network reciprocity, yet this conclusion typically assumes that individuals apply the same strategy toward all neighbors. Here, we relax this assumption and examine the evolution of prosocial punishment in a spatial Prisoner's Dilemma with both prosocial and antisocial punishers, allowing players to tailor their responses to each neighbor. This behavioral flexibility enhances prosocial punishment by enabling players to cooperate with cooperators and punish defectors. However, this advantage emerges only when behavior is updated through a balanced mix of direct experience and indirect social cues. Overreliance on either source undermines spatial reciprocity and results in lower prosocial punishment than models without flexibility. These findings demonstrate that the evolution of prosocial punishment depends not only on spatially constrained local interactions, but also on the ability to tailor responses across social ties and on the cognitive rules governing social learning, offering a potential explanation for why punishment often fails to evolve in experimental settings.