Andres Diaz-Pace - Agent-based Negotiation Techniques for improving Tradeoffs in Group Decision Making.

13:30
Mercredi
8
Juin
2016
Organisé par : 
Carole Adam
Intervenant : 
Andres Diaz-Pace, UNICEN (Tandil, Argentine)
Équipes : 
Information détaillée : 

Lieu : salle de séminaires du nouveau bâtiment IMAG

 

Andres Diaz-Pace is currently a professor at UNICEN University (Tandil, Argentina), and also an Independent Research Fellow of the National Council for Scientific and Technical Research of Argentina (CONICET). From 2007 to 2010, he was a member of the technical staff at the Software Engineering Institute (SEI, Pittsburgh, USA), with the Research, Technology, and System Solutions Program. His primary research interests are: quality-driven architecture design, AI and NLP techniques applied to software engineering, and architecture-based evolution and conformance. He has authored several publications on topics of design assistance and object-oriented frameworks. He has also participated, as an architecture evaluator or as a lead architect, in technology transfer projects with the software industry. Mr. Diaz-Pace received a Ph.D. in Computer Science from UNICEN University in 2004. Contact him at adiaz@exa.unicen.edu.ar

Résumé : 

Recommender systems that provide assistance to users (e.g., suggestions, alternatives, courses of action) for making decisions have received great attention over the last years, with applications to domains such as: tourism, movies, and recently software engineering tasks. When groups of users are involved, each user with her own goals and preferences, one important aspect of a (group) recommender system is how to archieve a balance (or tradeoff) that satisfies as much as possible all users´ goals. Along this line, we can model the problem as a multi-agent system (MAS), in which a personal agent assists each user, while the agents seek to collaborate to fulfill global goals of the group. In this context, we have investigated a negotiation technique called monotonic concession protocol (MCP) to work in the MAS approach above, so that the individual recommendations of the agents can be interchanged and negotiated towards improved tradeoffs among the recommendations. In particular, we have applied this approach in two different domains: movie recommendation for groups, and architectural design exploration based on different quality attributes. The seminar will cover the main concepts of our MAS and negotation-based approach, present some details and results of the two application domains, and finally wrap up with challenges and future research directions.