Projet de recherche doctoral numero :8321

Description

Date depot: 12 avril 2022
Titre: Explainable and Trustable Automated Negotiations
Directeur de thèse: Pavlos MORAITIS (LIPADE)
Encadrant : Jean-Guy MAILLY (LIPADE)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Intelligence artificielle

Resumé: In the coming 15 years artificial intelligence will have to participate in the digital transformation of the society by including new domains that are not currently accessible by the recent developments of (especially statistical) AI. These domains are related to different aspects of people's personal, social and professional life that generate conflicts which are time and money consuming for solving them. Such conflicts arise in divorces, road accidents, layoffs, business contracts, etc. Therefore, the aim of the forthcoming artificial intelligence research directions should be to facilitate people's personal and professional life by saving time and money. A research direction that may provide trustable solutions for conflicts resolution is automated negotiation. Apart the business domain where automated business negotiations should allow increasing profitability by saving time and resources as they could allow negotiating thousands of deals in parallel, we do believe that automated negotiation could also revolutionize other aspects of people's personal and social life by dealing for example with divorces, road accidents and layoffs. We call these negotiations legal negotiations. Indeed, the common characteristic of these three situations is that conflicts resolution is based on the strict application of a dedicated legislation that doesn't need a human judge's law subjective interpretation and sometimes personal intimate conviction as it is the case in other situations as for example homicides. Therefore, automating conflicts resolution involved in everyday people’s personal, social and professional life through automated negotiation will have an important social impact. However automated legal and business negotiations have to be explainable and trustable in order to be adopted by governments and people. That means that the agreements that will be proposed to the litigants (in the situations involving personal and professional people's life) have to be accompanied with explanations that present the laws (and the law cases) that have been applied, in order to create trust and allow acceptability. To the same way, the success and trust on automated business negotiations is based on the possibility of the negotiating parts to defend their offers (and counter offers) by explaining the reasons why an offer is proposed rather than another. Our aim in this thesis will be to propose a platform for automated legal and business negotiations based on computational argumentation [1] that allows to deal with the requirements presented above. This platform will allow automated argumentation-based legal negotiations for conflicts whose resolution is based on application of law (i.e. divorces, road accidents, layoffs) and automated argumentation-based business negotiations where conflicts resolution is based on the search of mutually accepted agreements. For realizing this work, we will extend the frameworks proposed in [2,3]. Indeed, the above frameworks are based on abstract argumentation and the challenge in this thesis will be to adapt them for structured argumentation as we will have to represent application specific knowledge for which the claims and the supporting evidence/information of the arguments must be explicitly described. For validating the theoretical work, we will be developing a proof of concept (POC) of the proposed platform. References [1] Y. Dimopoulos, P. Moraitis. Advances in Argumentation-based Negotiation, Chapter 4, in Negotiation and Argumentation in Multi-Agent Systems: Fundamentals, Theories, Systems and Applications, pp. 82-125, F. Lopes and H. Coelho (Eds.), Bentham Science Publishers, 2014 [2] Y. Dimopoulos, JG. Mailly, P. Moraitis. Arguing and Negotiating using Incomplete Negotiators Profiles, Autonomous Agents and Multi-Agent Systems, 35, 18 (2021) [3] Y. Dimopoulos, JG. Mailly, P. Moraitis. Argumentation-based Negotiation with Incomplete Opponent Profiles, in AAMAS'19, pp. 1252-1260, Montreal, Canada, 2019