Supporting Organizational Accountability inside Multiagent Systems

We present and analyze the problem of realizing an accountability-supporting system in multiagent systems technology. To this aim and to avoid ambiguities, we first characterize the concept of accountability, which we later work to realize computationally, particularly in relation to the partially overlapping notion of responsibility. Then, with the aim of achieving accountability as a design property, we provide a few principles that characterize an accountability-supporting multiagent system. We provide an accountability protocol to regulate the interaction between an agent, willing to play a role in an organization, and the organization itself. Finally, we show as a case study how the use of such a protocol allows multiagent systems, realized with JaCaMo, to support accountability.