Examine the filter bubble with a focus on emotion and content

Recommendation systems (RSs) have become popular in various Internet services, enabling the delivery of content tailored to user preferences. While RSs are designed to update recommended content to align more closely with user preferences based on browsing history and registration information, RSs may still present biased information due to inherent characteristics. This phenomenon is known as the "filter bubble,"a crucial aspect of the discourse on "informational health."Previous studies have examined the filter bubble phenomenon, focusing on the potential for bias in recommended content. However, there is no consensus on the definition of "bias,"and no agreement on the criteria for identifying a filter bubble state. In addition, the previous methods rely primarily on topic diversity as a metric, overlooking other important indicators. This study presents a more comprehensive approach to defining the state of the filter bubble. Our methodology employs two key indicators: content and emotion. We also evaluate the effectiveness of the proposed approach.