Accountability per i sistemi multi-agente: modellazione di un'accountability computazionale tra agenti software

Accountability offers strong advantages to organizations in multi-agent systems by providing a holistic picture after a situation occurs of who is effectively to blame/praise and in what capacity. Its dominion of inquiry stretches far beyond a simple cause/effect determination to include more complex ideas like mitigating circumstances and avoidance potential. The nuanced approach to situation analysis presents a great boon to systems seeking to provide a normative framework that justly sanctions its members. Accountability has two sides, one hard and one soft. Hard accountability refers to entities' adherence to their assigned tasks. Soft, on the other hand, concerns interference, namely actions that help/hinder organizational progress. As a continuance of our previous work, namely the ADOPT protocol, we propose an Object-Role Modeling model of hard accountability in order to graphically establish the information necessary to have an accountability relationship. Based on that relationship definition, we then establish accountability requirements for systems with precedence logic, followed by a Prolog-like definition of soft accountability.