WP0

Title: Organisation.

Leader: Stefano Moretti (LAMSADE).

Aim: Management of the project. The project is divided into five technical WPs, and a dissemination WP. The scientific coordinator of the project is Stefano Moretti (LAMSADE).

The scientific management of the project will be ensured by a board composed of the leader of the project and the leaders of the WPs. The partner coordinators are:

  • CRIL: Sébastien Konieczny

  • LAMSADE: Stefano Moretti

  • LIP6: Fanny Pascual

WP1

Title: Portfolio of solutions.

Leaders: Nicolas Maudet (LIP6), Meltem Öztürk (LAMSADE).

Aim: The main objective of this WP is to provide an answer to the general question of how to compare the elements of a finite set X given a ranking over the elements of its power-set (the set of all possible subsets of X). To do this, we use a classical methodology in Social Choice, Game Theory, and KR (Knowledge Representation and Reasoning): we propose a small set of properties that a social ranking solution should satisfy, and we prove that these properties identify a unique one. We want to develop a portfolio of solutions accompanied with a road-map of their properties aimed at driving a user toward the most adapted social ranking solution under a given scenario. The axiomatic analysis of solutions plays a central role in this WP, together with the computational issues related to the associated algorithms.

WP2

Title: Subdomains, elicitation and manipulation.

Leaders: Paolo Viappiani (LIP6), Denis Bouyssou (LAMSADE)

Aim: How to choose axioms to characterize a social ranking over particular sub-domains (e.g., partial orders, single-peaked coalitional relations, ranking over coalitions with the same cardinality, etc.) constitutes the main issue for this WP. Another objective is to better understand the consequences of the incomparability of certain coalitions, that may arise for many practical reasons (for instance, incompleteness of the data, heterogeneity of criteria, lack of information about the nature of the coalitions, etc.). Considering sub-domains has important implications for the definition of social recommendation in the sense that we will be interested in studying incremental elicitation problems where we ask questions about pairwise comparisons on coalitions aimed at finding the outcome of a social ranking solution as fast as possible. In these domains, we are also interested in analysing the robustness to manipulation of different solutions, and its impact on the strategic behaviour of individuals.

WP3

Title: Coalition formation.

Leaders: Nawal Benabbou (LIP6), Laurent Gourvès (LAMSADE) and Fanny Pascual (LIP6)

.

Aim: We will investigate the effect of social ranking solutions on the behaviour of individuals to form coalitions: is it better to cooperate in larger or smaller coalitions? Is it convenient for two or more coalitions to merge? Is a solution resistant to the manipulation of individuals, who could strategically affect the position of groups in the ranking? These questions are strongly related to the dynamics of coalition formation in hedonic games and related models where the preferences of agents are extracted from an exogenous “power” relation over the set of coalitions. These aspects will also be analysed in connection with algorithmic issues arising from specific interaction dynamics.

WP4

Title: Compact representation.

Leaders: Patrice Perny (LIP6) and Sébastien Konieczny (CRIL).

Aim: A critical aspect that must be considered for the algorithmic design of alternative solutions is the high computational burden generated by the pairwise comparison of an exponential number of possible coalitions. Dealing with ordinal information about coalitions and the complexity related to the computation of social ranking solutions, we face the interesting problem of merging models from the literature about compact representation of cooperative games with those about compact preference representation. As a complementary approach, we will also study disaggregation models based on numerical representations of preference orders that will facilitate the formulation of the social ranking problem as a combinatorial optimization problem.

WP5

Title: Explaining ordinal influence in social AI.

Leaders:Stefano Moretti (LAMSADE), Paolo Viappiani (LIP6).

Aim: In order to show the generality of our framework, and its versatility over different domains, we will apply our solutions to evaluate the influence of criteria for the selection of students in the national admissions platform Parcoursup, analysing governmental datasets that are freely available on-line[1]. Within this application framework, we will focus on the following objectives:

  • studying a classification model to reproduce the outcomes of Parcoursup during the previous years;

  • applying our portfolio of social ranking solutions to evaluate the relevance of different criteria for the selection of students and to explain the prediction over the classes established by Parcoursup.

If project time allows, our portfolio of solutions will also be tested over alternative domains of application: in social network analysis, to evaluate and compare the ordinal influence of agents in social networks and to evaluate the positive and negative synergies between criteria in MCDA.

WP6

Title: Dissemination, Popularization, Education.

Leaders: Meltem Öztürk and Gabriella Pigozzi (LAMSADE).

Aim: The aim of this WP is to promote the results of the project at the academic, educational, and popularization level. We dedicate a budget for this dissemination task.