文摘
Norms are commonly understood as guides for the conduct of autonomous agents, thereby easing their individual decision-making and coordination. However, their study exhibits a polarity between (i) norms as behavioural patterns emerging from repeated agents-(inter)actions and (ii) norms as explicit prescriptions. In this paper, we attempt to build a bridge between these two conceptual poles of norms: it takes the form of a mental function for prescriptive transfiguration allowing reinforced learning agents to express their learning experiences into prescriptions. The population of transfigurative agents are then equipped with a consensus system to build and enforce prescriptive systems to self-govern on-line. Simple simulations suggest the pertinence of the approach and shows its weaknesses, in particular prescriptions stalling learning, and timeliness in norm construction.