Enabling cross-continent provider fairness in educational recommender systems
With the widespread diffusion of Massive Online Open Courses (MOOCs), educational recommender systems have become central tools to support students in their learning process. While most of the literature has focused on students and the learning opportunities that are offered to them, the teachers be...
| Autores: | , , , , |
|---|---|
| Tipo de documento: | artigo |
| Estado: | Versión aceptada para publicación |
| Data de publicação: | 2022 |
| País: | España |
| Recursos: | Universidad de Barcelona |
| Repositório: | Dipòsit Digital de la UB |
| OAI Identifier: | oai:diposit.ub.edu:2445/194424 |
| Acesso em linha: | https://hdl.handle.net/2445/194424 |
| Access Level: | Acceso aberto |
| Palavra-chave: | Intel·ligència artificial Cursos en línia oberts i massius Artificial intelligence Massive Open Online Courses |
| Resumo: | With the widespread diffusion of Massive Online Open Courses (MOOCs), educational recommender systems have become central tools to support students in their learning process. While most of the literature has focused on students and the learning opportunities that are offered to them, the teachers behind the recommended courses get a certain exposure when they appear in the final ranking. Underexposed teachers might have reduced opportunities to offer their services, so accounting for this perspective is of central importance to generate equity in the recommendation process. In this paper, we consider groups of teachers based on their geographic provenience and assess provider (un)fairness based on the continent they belong to. We consider measures of visibility and exposure, to account () in how many recommendations and () wherein the ranking of the teachers belonging to different groups appear. We observe disparities that favor the most represented groups, and we overcome these phenomena with a re-ranking approach that provides each group with the expected visibility and exposure, thus controlling fairness of providers coming from different continents (cross-continent provider fairness). Experiments performed on data coming from a real-world MOOC platform show that our approach can provide fairness without affecting recommendation effectiveness. |
|---|