Furando a Bolha: Nudges Digitais em Sistemas de Recomendação

• Gabrielle Alves USP
• Marcelo G. Manzato USP

DOI: https://doi.org/10.5753/webmedia_estendido.2024.244374

Resumo

This research explores the challenges of recommendation systems, focusing on diversity and bias mitigation. Excessive precision can create filter bubbles, limiting users’ worldviews and reinforcing existing preferences – a path to injustice and polarization. We investigate how nudge mechanisms can encourage interactions with diverse content, reducing biases. We analyze user interactions with recommendations, both with and without these nudges, using qualitative and quantitative methods. The results show that nudges increase diversity without compromising system quality. This research deepens the understanding of how to integrate diversity into recommendations while maintaining user satisfaction and suggests practical improvements to enhance fairness and improve the user experience.

Palavras-chave: Sistemas de recomendação, diversidade, vieses, nudges, bolhas de filtro, justiça

Referências

Ricardo Baeza-Yates. 2018. Bias on the web. Commun. ACM 61 (05 2018), 54–61. DOI: 10.1145/3209581

Jiawei Chen, Hande Dong, Xiao lei Wang, Fuli Feng, Ming-Chieh Wang, and Xiangnan He. 2020. Bias and Debias in Recommender System: A Survey and Future Directions. ArXiv abs/2010.03240 (2020).

Dietmar Jannach, Markus Zanker, Alexander Felfernig, and Gerhard Friedrich. 2010. Recommender Systems: An Introduction. Recommender Systems: An Introduction (01 2010). DOI: 10.1017/CBO9780511763113

Mathias Jesse and Dietmar Jannach. 2021. Digital Nudging with Recommender Systems: Survey and Future Directions. Computers and Human Behavior Reports 3 (2021).

Thorsten Joachims. 2002. Evaluating Retrieval Performance Using Clickthrough Data. (2002), 79–96.

Eli. Pariser. 2011. The filter bubble : what the Internet is hiding from you. Penguin Press, New York.

Pearl Pu, Li Chen, and Rong Hu. 2011. A User-centric Evaluation Framework for Recommender Systems. In Proceedings of the 5th ACM Conference on Recommender Systems (RecSys ’11). 157–164.