Decision making in multi-agent settings is a complex exercise where agents have to handle incomplete knowledge of the complete problem. Agents are interdependent in multi-agent decision making, being subject to the decisions of other agents who bring to bear other qualitative and quantitative criteria. Some aspects of this problem have been addressed in the Distributed Constraint Optimisation Problems (DCOP) and Markov Decision Processes literature. Taking inspiration from a medical example, our objective in this paper is to provide a framework to support multi-agent decision coordination. This method can be applied in scenarios where we seek to combine qualitative preferences on projected final states with assessment made using utility/objective functions, while accounting for partial agent knowledge.
LONDON - March 31, 2022 Highlights: Česká spořitelna has been able to reduce collector call minutes by 25%, while maintaining a stable. | March 30, 2022
The choices we make constrain other choices. Certain combinations of decisions lead to better outcomes than other combinations of decisions. These observations hold in a large class of problem domains, but are especially compelling in policing and law enforcement. However, the problem of identifying the optimal choices has received relatively little attention in the literature. The nearest point of departure is the class of Distributed Constraint Optimisation Problems (DCOP), but these offer a limited vocabulary for describing the final outcome (we are only able to assess the value of a shared objective function when a complete assignment of values to all variables of interest has been arrived at). In many settings, the final state of affairs that is achieved after deploying a sequence of decisions is a critical determinant of the effectiveness of the choices made. To address these gaps in the literature, we introduce a novel notion of a Decision Interdependency Network (DIN). We show