Description du projet
Vers des systèmes de recommandation de groupe intelligents, fiables et transparents
Les écosystèmes de commerce électronique et les plateformes de médias sociaux utilisent des outils intelligents pour les recommandations de contenu et la modération des messages. Les systèmes de recommandation de groupe (SRG) gagnent en popularité, mais la demande d’une plus grande transparence des algorithmes et des processus de décision a un impact négatif sur les opportunités d’emploi, le commerce électronique, ainsi que sur l’exposition aux actualités. Il est essentiel de garantir l’impartialité et l’explicabilité pour établir la transparence et renforcer la confiance dans les systèmes basés sur l’IA tels que les SRG. Toutefois, si les systèmes de recommandation (SR) individuels traditionnels se sont efforcés d’améliorer ces aspects, ils n’ont pas encore été étudiés dans le cadre des systèmes de recommandation généraux. Le projet FIDELITY, financé par le programme MSCA, se consacrera au développement de nouveaux algorithmes et outils pour améliorer l’explication, l’impartialité et la synergie au sein des SRG. Le projet intégrera des techniques d’explication post hoc pour combler le fossé entre l’explication et l’impartialité dans les SR et les SRG.
Objectif
Today, most social media networks use automated tools to recommend content or products and to rank, curate and moderate posts. Recommender systems (RSs), and in particular Group recommender systems (GRSs), -a specific kind of RSs used to recommend items to a group of users-, are likely to become more ubiquitous, with expected market forecast to reach USD 16.13 billion by 2026.
These automated content governance tools are receiving emerging interest as both algorithms and decision-making processes behind the platforms are not sufficiently transparent, with a negative impact on domains such as fair job opportunities, fair e-commerce or news exposure.
Two of the key requirements that have to be fulfilled to build and keep users’ trust in AI systems while guaranteeing transparency are Fairness and Explainability. But, aside from some previous attempts to enhance both aspects in traditional-individual RS, they have hardly been explored in GRSs.
FIDELITY addresses this challenge by developing novel algorithms and computational tools in GRS to boost explanation, fairness, and synergy between them through a disruptive multidisciplinary research approach that: 1) extensively brings SHAP and LIME, as state-of-the-art post-hoc explanation approaches in AI, into RS and GRS contexts, 2) bridges explanation and fairness in RS and GRS, introducing an explanation paradigm shift moving from “why are the recommendations generated?” to “how fair are the generated recommendations?” and, 3) transversally evaluates the new methods through real-world GRSs and user studies. The ultimate goal is to guarantee greater user trust, and independence of RS output from any of the sociodemographic characteristics of users. The training programme, designed with the aim to fill the existing gaps between computing science, social research and business development reality, will provide the candidate with a multidisciplinary background that will boost his innovation potential and career prospects.
Champ scientifique
Programme(s)
- HORIZON.1.2 - Marie Skłodowska-Curie Actions (MSCA) Main Programme
Régime de financement
HORIZON-TMA-MSCA-PF-EF - HORIZON TMA MSCA Postdoctoral Fellowships - European FellowshipsCoordinateur
23071 Jaen
Espagne