VAlorisation des DOnnées pour la Recherche d’Emploi

We study the properties and the implementation of a job recommender system at Pôle emploi. This system is able to recommend job ads to job seekers and, symmetrically, job seekers to job ads. Its development was made possible by extensive work from our multi-disciplinary research team (data-scientists and economists) that has lasted for four years. Our project already received substantial funding from DataIA in 2018, which has allowed us to develop our algorithms as well as the necessary pipelines to implement them at Pôle emploi. The final stages of this vast project remain to be completed: namely, the implementation of Randomized Controlled Trials (RCTs) to measure the impacts of the recommender system on job search behavior, recruitment procedures and return to employment, the finalization of the algorithms and the writing of academic papers in both data-science and economics. The final stages of our project will also include the development of an extension of our recommender system that aims at controlling congestion in the labor market (i.e. the fact that many job seekers are shown the same few vacancies, and vice versa), using tools from the computational optimal transport literature. In order to complete this project, we are requesting funding for an additional year of PhD for one of our students, as well as funds to hire a data-science research engineer for one year.