R.A. Flores (USA) and R.C. Kremer (Canada)
Multiagent systems, Autonomous systems, Conversations.
Interactions in multiagent systems are traditionally defined using conversation protocols. More often than not, agent designers build these protocols without following any prin ciples defining the sequencing of messages and the turn taking for their utterance. On the one hand, by not be ing burdened by methodological constraints, designers can effortlessly create protocols using any message sequences they believe are best to achieve the goals of their systems. On the other hand, these protocols are useful for as long as agents do not attempt to diverge from them, which in the case of systems with agents of various levels of contextual responsiveness would restrict any conversational optimiza tions that these agents may attempt. To bridge the ability of creating protocols for simple agents, while supporting the conversational flexibility required by sophisticated agents, we propose an inference model where the sequencing and turn-taking in conversation is guided by negotiations to ad vance the state of social commitments. The main contribu tion of this model are to allow the definition of protocols in a principled manner and to dictate the inferential evo lution of conversations, while supporting the heterogeneity and flexibility required by agents in open environments.
Important Links:
Go Back
