Multi-Agent Systems are suitable to provide a framework that allows to perform collaborative processes in distributed environments. Furthermore, argumentation is a natural way of reaching agreements between several parties. We propose an infrastructure to develop and execute argumentative agents in an open MAS. It o ers the tools to develop agents with argumentation capabilities. It also o ers support for agent societies and their social context, which allows agents to engage in argumentation dialogues in more realistic environments. In our application scenario, the argumentative agents try to reach an agreement about the best solution to apply to solve a problem reported to the system.
The main components of our infrastructure are the argumentative agents, the Commitment Store and the knowledge interchange mechanism.
The argumentative agents have all the tools needed (argumentation protocol,
communication skills and knowledge resources) to engage in an argumentation
dialogue and reach an agreement with other agents about the best solution to
apply for a problem. The solution applied to solve a problem in the past and
the information about the problem-solving process can be reused to propose a
solution to another similar problem. Case-Based Reasoning (CBR) systems have
been widely applied to perform this task [
The Commitment Store is a resource that stores all the information about the agents participating in the problem-solving process.
The case-bases of the domain CBR and the argumentation CBR are stored as OWL 2 data of an ontology that we have designed to act as language representation of the cases. In this way, heterogeneous agents can use it as common language to interchange solutions and arguments generated from the case-bases.
Furthermore, the proposed infrastructure has been validated with an example in a customer support application. In this example, the agents' social context simulates the roles of operators, experts and managers attending incidences of users. There are dependency relations between the roles representing a hierarchy. Also, each agent of the system has its own preference values to choose the solution to apply to a problem. In the performed tests, the best results are obtained using an argumentation policy that takes into account the social context of agents. In addition, having at least an expert involved in the group of agents that tries to solve a problem increases the quality of the nal solution agreed. 3
In this work, we have implemented an infrastructure to develop and execute argumentative agents in an open MAS. This infrastructure o ers the necessary tools to develop agents with argumentation capabilities, including the communication skills and the argumentation protocol. Also, it o ers support for agent societies and takes into account the agents' social context. The infrastructure combines the CBR methodology, argumentation and MAS.