Recomendado Para Você: o impacto do algoritmo do YouTube na formação de bolhas
The emergence of algorithmic curation systems on the web that customizes webpages, showing only content that the user is interested in has generated several concerns among researchers, especially about filter bubbles: an environment where there is only one predominant point of view, and diverging op...
| Autor: | |
|---|---|
| Tipo de documento: | dissertação |
| Estado: | Versão publicada |
| Data de publicação: | 2018 |
| País: | Brasil |
| Recursos: | Universidade Federal de Minas Gerais (UFMG) |
| Repositório: | Repositório Institucional da UFMG |
| Idioma: | português |
| OAI Identifier: | oai:repositorio.ufmg.br:1843/BUOS-B6GEZC |
| Acesso em linha: | http://hdl.handle.net/1843/BUOS-B6GEZC |
| Access Level: | Acceso aberto |
| Palavra-chave: | Algoritmos Bolha dos filtros Balcanização YouTube Comunicação YouTube (Recurso eletrônico) |
| Resumo: | The emergence of algorithmic curation systems on the web that customizes webpages, showing only content that the user is interested in has generated several concerns among researchers, especially about filter bubbles: an environment where there is only one predominant point of view, and diverging opinions are filtered out. However, there are studies that point that the web tends to increase the diversity of points of view, as they allow each user to keep in touch with a higher number of friends. We propose a study of the impact of the algorithmic recommender system inside YouTube to observe if it is responsible for increasing or reducing diversity, and if there is a filter bubble caused by it. To analyze this phenomenon, we propose a study using different user profiles created specifically for this research, one simulating a user leaning politically to the left, and another one leaning politically to the right, observing the recommendations made to each one by the algorithm, and the diversity of them, allowing us to observe whether there is a reduction of videos that are presented to both or a polarization caused by the algorithm. We made a data collection on two different subjects, and later an analysis of the data obtained in two parts, first with a more qualitative approach, observing the pages, the order of recommendations, thumbnails and titles, and finally we made a network analysis, which allows us to get a broader view of the recommendations generated by algorithm. We have found that there isnt a clear pattern that can show us if there is or not a formation of a filter bubble, but there is a reduction of diversity, however it occurs in relatively different ways depending on the theme. This means that it is possible to observe a reduction of the diversity that causes a bubble, but we need to be cautious when affirming that the algorithms creates a filter bubble in a specific case. It is very important to analyze each individual scenario to be sure that there is indeed a decline in diversity that can lead to a filter bubble. |
|---|