Advanced Multilateral Argumentation for Deliberation – AMANDE

formal argumentation, logic, game-theory, agent-based simulations

- survey on aggregation of multiagent systems
- survey on gradual semantics
- new approaches for judgement aggregation, based on vote support
- new approaches for gradual argumentation semantics
- new operators for argumentation dynamics
- encoding in logic of dynamic operators (which paves the way for very efficient solving methods)

New argumentation semantics, study of argumentation dynamics, detecting manipulation

Scientific publications

The AMANDE (Advanced Multilateral Argumentation for DEliberation) project aims at providing formal tools for designing multiparty deliberation systems. The main objective is to define and study the properties of such systems, so that designers can ensure that users and possibly interacting agents are able to identify key arguments in a deliberation process, and to fairly defend their view. While the technological tools are now available to deploy such systems (see last paragraph), we believe there is now an urgent need to work on the theoretical foundations of such platforms.

Argumentation is at the heart of deliberation processes since people generally justify their positions and opinions by arguments. Since the early nineties, argumentation theory is studied in AI. The main limitation of existing works on argumentation theory is that they are mostly concerned by a unique agent carrying on an argumentation/deliberation process from its available information (even if some works use the metaphor of dialectical proofs). Our aim is to define a genuine multilateral theory of argumentation, involving potentially many agents, where each agent has her own information (arguments, beliefs, goals, etc.). For that purpose, we will take advantage of different formal tools, especially those developed in economy for modeling group interaction and decision processes, such as social choice (vote), game theory, and also works on logical aggregation such as belief merging and judgment aggregation. We will study how to define a common point of view regarding an argument from a set of argumentation systems. This will require to develop specific methods, but more importantly to define and study the expected properties of such methods. Once this notion of rational output will be understood, it will be important to try to distribute the process to obtain a more realistic debate framework with autonomous agents. Then the questions will be to know whether such rational output can be reached in the distributed case and which are the obtained rational properties. We shall also consider individual strategic behaviour (especially in the perspective of implementing autonomous entities), and we will in particular study if and how an agent can manage to manipulate the deliberation process. Similarly we shall consider group behaviours and study how coalition of agents can emerge.

More precisely, our main ambition with this project is to provide the theoretical foundations allowing a designer to :
1. support multiparty deliberation, allowing for instance to fairly regulate a debate and avoid manipulation, to assess the status of a given claim, or to emphasize the critical claims put forward so far.
2. analyze multiparty deliberation, allowing for instance to identify key arguments, the most influential contributors, or the possible coalitions.
3. enable artificial agents to effectively take part in such deliberations, which raises in addition questions pertaining to the adequate definitions of strategies for such agents.

An important application with great potential impact is related to debate systems that are emerging on the web (see for instance Debatepedia (http://debatepedia.idebate.org/) and DebateGraph (http://debategraph.org/). The success of these platforms in their currrent form seems to suggest that they can become an important source of information, just as wikipedia is now. For instance, DebateGraph was used to produce maps for the British newspaper “The Independent” and talk shows on CNN, and is supported by the White House, the European Commission, and other institutions. These debate systems are still in their infancy though. For the moment they are mainly interfaces where people can give arguments pro or con a given issue without any particular processing and evaluation of those arguments. The AMANDE project will provide automatic reasoning/decision capabilities to these platforms.