Bridging the Specification Protocol Gap in Argumentation

Abstract

Coordinating agents in open environments is a difficult problem that has engaged mul-ti-agent systems researchers on three broad fronts: argumentation (which focuses on the basis for negotiation between agents); electronic institutions (which focuses on the norms of social interaction in agent groups) and protocols (which focuses on deployment of interactions). None of these areas subsumes the others but there is a strong interaction between them in many cases. For example, if one wishes to construct a multi-agent trading system then this will contain negotiation, restrictions on group behaviour and protocols relevant to each agent. This is broadly analogous to the relationships between different views of formal system requirements in software engineering. Accordingly, the interesting challenge here is how (or whether) the specification of one of these components can be used to constrain the specification of the others. One way of addressing this issue is through the automated synthesis method, so the specific question that we ask is whether a generic argumentation representation (acting as a high level specification language) can be used to automate the synthesis of executable specifications in a protocol language capable of expressing a class of multi-agent social norms. As our argumentation language we have chosen the Argument Interchange Format (AIF is a generic specification language for argument structure). As our protocol language we have chosen the Lightweight Coordination Calculus (LCC is an executable specification language used for coordinating agents in open systems). Fully automated synthesis starting only from the AIF is not possible because AIF is an abstract language that does not capture some concepts that are related to the interchange of arguments between agents (e.g. sequence of argument, locutions and pre-and post-conditions for each argument). An example of this problem occurs in one of the basic dialogue games stereotypes: A1 and A2 are reasoning about whether a particular penguin, Tweety, can fly: A1) Tweety flies. (making a claim); A2)Why does Tweety fly? (asking for grounds for a claim); A1) Tweety is a bird, birds generally fly. (arguing: offering grounds for a claim); A2) Tweety does not fly because Tweety is a penguin, penguins do not fly. (stating a counterargument); A1)You are right. (conceding an argument). In this dialogue game, each agent responds in turn to the argument made by other agent. This flow of the dialogue is not captured by AIF (e.g. AIF does not record that a given argument has been made in response …

Cite this paper

@inproceedings{Maghraby2012BridgingTS, title={Bridging the Specification Protocol Gap in Argumentation}, author={Ashwag Maghraby and Dave Robertson and Adela Grando and Michael Rovatsosr}, year={2012} }