Situated artificial institutions: stability, consistency, and flexibility in the regulation of agent societies

International audienceIn multi-agent systems, norms are a usual way to regulate the behaviour of autonomous agents. To be stable in different circumstances, norms are specified using high level terms, abstracting from the particular dynamics of the environment where the agents are situated. However, applying these norms requires a proper link with a concrete environment. Detaching that link from the norms themselves provides stability to the normative regulation but raises consistency and flexibility issues. Consistency is achieved when the abstract norms are coherent with the environment under regulation. Flexibility is achieved when different kinds of norms share the same interpretation about the environmental state. These properties are provided in some current works. However, since they are interrelated, there is not, to our knowledge, a single proposal providing all of them. This paper proposes the situated artificial institution (SAI) model to address these three issues—stability, consistency, and flexibility—by conceiving norms as part of institutions that provide, through the process of constitution, a social interpretation of the environmental state. After the presentation of the formalised model of SAI, a case study is used to illustrate and test this approach