SSLE: A framework for evaluating the “Filter Bubble” effect on the news aggregator and recommenders
Han Han, Can Wang, Yunwei Zhao, Min Shu, Wenlei Wang, Yong Ki Min
Abstract
Abstract Recommendation algorithms are data filtering tools that make use of algorithms and data to recommend the most relevant items to a particular user. The algorithm-driven recommenders become indispensable and supersede search engines as the most important information dissemination channel. On one hand, it becomes an integral component in the existing social media, e.g. Weibo, Twitter, etc. On the other hand, news aggregators and recommenders have proliferated and gained an increasing market share. As a result, the previous studies usually study the “filter bubbles” phenomenon in the context where the social filtering dominates the dissemination of information. However, less attention is paid to the news aggregators and recommenders where algorithm-driven technological filtering dominates. Therefore, in the previous research, “filter bubbles” are usually equated with the community structure, but lack of the detailed analysis of the content agglomeration through the users’ interaction with the platforms. Based on these concerns, we propose a four-phase (“Selection”, “Setup”, “Link”, and “Evaluation”) skeletal solution framework targeted at exploiting the filter bubble effect of the personalized news aggregation and recommendation system. Furthermore, we illustrate the effectiveness of the proposed framework with a case study in three top Chinese news aggregators, i.e. Toutiao, Baidu News, and Tencent News. The results show that the users are narrowed into one or a limited number of topics over time. The phenomenon of the narrowed topics is deemed as the emergence of the “filter bubbles”. We also observe that the filter bubbles demonstrate different convergence degrees as user’s individual preference varies.