Role-based Semantics for Agent Communication Languages

We consider agents having multiple communication sessions at the same time. We assume that FIPA semantics of agent communication languages can still be used when we attribute mental attitudes for each session, which we call the roles of the agents, and we assume that we have to distinguish the mental attitudes attributed to the roles from the mental attitudes of the agents. We consider several consequences of the distinction between the mental attitudes attributed to the roles and the mental attitudes attributed to the agent. First, in attributing mental attitudes to an agent or to one of its roles, we argue that only mental attributes are attributed to an agent’s role when these attributes follow directly from the agent’s communication. They are therefore public in the sense that every agent who has overheard the session, has the same beliefs about the mental attitudes of the role. Second, the moves available to the dialogue participants in the same dialogue game are based on the role only, such that different kind of moves can be specified in different types of dialogue games. Obligations are associated to roles related to institutions which can enforce them by means of sanctions. Third, expectations are based both on the mental attitudes ascribed to the agent and to the role. 1 Multiple sessions at the same time Consider an agent who informs another agent that his web services can be used by all other agents, informs a third agent that his services can only be used by agents with the appropriate certificates, requests from a fourth agent a document, and informs a fifth agent that he does not have a goal to obtain this document. The semantics of speech acts in mentalistic approaches like FIPA [8], specified in terms of plan operators whose preconditions refer to the beliefs, goals and intentions of agents, cannot model such an insincere agent. In this paper we therefore generalize the FIPA model for the various sessions or dialogues of such an agent. 2 Mental attitudes for each session Though an agent may tell two agents incompatible stories, it seems much less useful that an agent is allowed to tell incompatible stories to the same agent. We therefore assume that FIPA semantics of agent communication languages can still be used, when we attribute mental attitudes for each session. We call the mental attitudes of an agent associated with such a session a role of the agent, for reasons that will become apparent only later in this paper. In the literature, the use of roles within agent communication is controversial. On the one hand, communication among agents in a multiagent system is often associated with the roles agents play in the social activity that is automated by the system. The GAIA methodology for agent-based software design [16] proposes interaction rules to specify communication among roles, the ROADMAP methodology [10] specifies in a so called social model the relations among roles, and in AALAADIN [7] interaction is defined only between the roles of a group: “The communication model within a group can be more easily described by an abstracted interaction scheme between roles like the ‘bidder’ and the ‘manager’ roles rather than between individual, actual agents”. On the other hand, most approaches to the semantics of agent communication languages do not take into account the fact that communication always takes place among agents in a role. Role names, like ‘speaker’ and ‘addressee’ or ‘buyer’ and ‘seller’ are often mentioned in the definition of agent communications languages. However, these terms only serve to bind individual agents to the speech acts in the protocol, but they are not associated with a state which changes during the conversation. The function of roles in dialogue we study in this paper is similar to the function they play in an organization, where they define the power of agents to create institutional facts, like commitments. As in organizations, it is possible that the same agent plays different roles, thus determining ambiguities and conflicts. For example, a command issued by a friend may not be effective, unless the friend is also the addressee’s boss. 3 The agent versus its roles We assume also that we have to distinguish the mental attitudes associated with the roles from the mental attitudes of the agents. For example, if we can inspect the knowledge base of an agent, then we should attribute the knowledge not to one of its roles, but to the agent itself. Moreover, we need to represent the mental attitudes of the agent itself if we wish to model that the agent is lying. Therefore, our model of a dialogue between two agents can be visualized as in Figure 1. The circles x and y represent two agents playing respectively roles r1 and r2 in the dialogue game. We use indices i, j to range over role instances, so in the figure we have i = x : r1, j = y : r2.