Agent Support for Collaboration in Complex Deliberative

Organisations often form collaborative teams for sharing knowledge and resources in order to accomplish complex tasks. For example, a team involving non-governmental organisations, civil and military institutions may be formed rapidly in response to humanitarian crisis. In another context, a company may collaborate with a team of service providers in fulfilling its requirements. Collaboration enhances the capabilities of individuals. However, working in a team requires a great effort from the individuals to overcome to differences in, for example, knowledge, instruments, and culture. Incompatible objectives in addition to internal policies increase the complexity of cooperation in decision-making. For effective teamwork, team members engage in discussion to agree on a common plan of activities, but often positive outcomes are hampered by conflicts of opinion among partners on the way to proceed. Software agents, employing mechanisms that support a discussion regarding a shared plan, have the potential to facilitate conflict resolution among partners. We believe that the use of argumentation-based models to support dialogues would aid team members to identify effective justifications and more favourable agreements. Recent work on argumentation for deliberative dialogues in multi-agent system has shown that argumentation is a promising approach for generating consistent collaborative plans [1, 2]. On the other hand, some earlier research has exploited agent support for policy advice in producing more cost/time efficient plans within a team planning context [7]. This existing research, however, does not adequately address the requirements of applications where agents are concerned with agreeing interdependent plans. For example Atkinson and Bench-Capon [1] present an argumentationbased dialogue for practical reasoning based on argumentation schemes where agents focus on the discussion of what is possible and justified in a shared plan with a common objective, according to values or costs. In contrast, in this research we explore what is possible when agents elaborate individual plans for achieving different objectives where only some activities require cooperation. Furthermore, plan constraints are not sufficiently expressive for describing the reasons for adopting a certain plan. A team member may