This work is related to the collaboration between the SMAC team (Multiagents system and behaviour) of the LIFL (Computer Science Laboratory of Lille) and the Applied Intelligent Systems Laboratory at University of Applied Sciences Western Switzerland, Sierre (HES-SO).
The work will be performed within the Applied Intelligent Systems Laboratory:
University of Applied Sciences Western Switzerland, Sierre (HES-SO)
TechnoArk 3
3960 Sierre, Switzerland
Argumentation can be abstractly defined as the interaction of different arguments for and against some conclusion. Over the last few years, argumentation has been gaining increasing importance in multiagents systems, mainly as a vehicle for facilitating "rational interaction" (i.e., interaction which involves the giving and receiving of reasons). This is because argumentation provides tools for designing, implementing and analysing sophisticated forms of interaction among rational agents.
In this perspective, the ArguGRID project (in which S. Bromuri and M. Morge have participated) has providen an agents environment for service-oriented computing [1]. For this purpose, software agents argue in bilateral negotiation for selecting and composing services. The ArguGRID platform is built upon a bilateral negotiation framework (reasoning mechanism, negotiation strategy, communication protocol) .
The aim of this work is to introduces a simple and intuitive agents environment for multiparty argumentation, in which several (more than two) agents are equipped with argumentation systems.
Argumentation space. The theory of tuple spaces is a model of communication between distributed computational entities. The essential idea is that computational agents connected together may create named object stores, called tuples, which persist until explicitly deleted. The first objective is to build such a blackboard-like shared data store for arguments which is a public-write and public-read space. In this way, any agent in the system may push a new argument and any agent may contradict an existing argument.
Argumentation protocol.
What we need to show is how two-person immediate response protocols can be generalized and extended to a multi-agent setting by means of a simple blackboard metaphor. For this purpose, we need to look at the way turn taking is organized in the system and how the winner is determinated. Moreover, we hope to start looking into properties such as the termination, the soundness and the completeness of the dialogues [2].Argumentation reasoning.
Obviously, the agent reasoning [4] need to be adapted to multiparty dialogues. In particular, the dialogue strategies of agents must take into account the dialogue protocol and the structure of the argumentation space.Heath System.
We plan to build upon our agents environment a health system where an agent diagnosis diabetes [3] and the symptoms can be confirmed by a second or a third expert. We can consider the following scenario. A diabetes expert, given glucose and blood pressure and pulse, may reach the conclusion that a hypoglicemia is taking place. A set of other things that may take place in the patients involves, related to hypoglicemia: transient ischemic attack, stroke, miocardial infarctuation which involves though the confirmation of an arrythmia. Consequently the diabetes expert may ask for confirmation of a symptom, given the existing ECG of the patient, to a second expert, which now is capable of detecting arrythmia. If the arrythmia is confirmed, then an alert is fired by the first agent stating that there is a high risk of bad complications, to a medical doctor, which then has to overview the report generated by the experts, that includes hypoglicemia plus one of the other bad conditions (according to the type of arrythmia detected).A grant (more than 421 euros/month) will be providen by the Applied Intelligent Systems Laboratory