Amplification pipelines: the role of feedback loops in recommender system bias

• Ano de defesa: 2023
• Autor(a) principal: Lente, Caio Truzzi
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Algorithmic bias; Aprendizagem de máquina; Machine learning; Recommender system; Sistemas de recomendação; Viés algorítmico
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/45/45134/tde-25032024-124902/

Recommendation algorithms have become essential to various systems we use on a daily basis, from what movies to watch to what products to buy. However, with the proliferation of these models on social networks, new concerns have come to light. Anecdotal evidence and an ever growing body of research indicate that social network algorithms that promote engaging content might be radicalizing users through the amplification of fringe viewpoints. The present study aims to examine recommendation algorithms dynamically as a means to identify feedback loops that could end up polarizing and radicalizing users.