A study on the termination of negotiation dialogues

Dialogue represents a powerful means to solve problems using agents that have an explicit knowledge representation, and exhibit a goal-oriented behaviour. In recent years, computational logic gave a relevant contribution to the development of Multi-Agent Systems, showing that a logic-based formalism can be effectively used to model and implement the agent knowledge, reasoning, and interactions, and can be used to generate dialogues among agents and to prove properties such as termination and success. In this paper, we discuss the meaning of termination in agent dialogue, and identify a trade-off between ensuring dialogue termination, and therefore robustness in the agent system, and achieving completeness in problem solving. Then, building on an existing negotiation framework, where dialogues are obtained as a product of the combination of the reasoning activity of two agents on a logic program, we define a syntactic transformation of existing agent programs, with the purpose to ensure termination in the negotiation process. We show how such transformations can make existing agent systems more robust against possible situations of non-terminating dialogues, while reducing the class of reachable solutions in a specific application domain, that of resource reallocation.