Viblio: Introducing Credibility Signals and Citations to Video-Sharing Platforms

Title of the Paper

Viblio: Introducing Credibility Signals and Citations to Video-Sharing Platforms

Paper Information

• Subject Area: Content credibility and misinformation governance on social media platforms
• Keywords: Credibility signals, citations, YouTube, misinformation, social platforms, contextual inquiry, semi-structured interviews

Research Background and Problem

• Problem or Challenge: On video-sharing platforms (e.g., YouTube), users are easily exposed to false or misleading information, while effective tools for assessing credibility are lacking. Although existing research focuses on misinformation detection, there is less attention on how users evaluate content credibility and how to improve these methods.
• Significance: Video-sharing platforms are becoming a primary source of information for many people, and their recommendation algorithms may exacerbate the spread of misinformation. Young users increasingly rely on social platforms for information, necessitating the establishment of more reliable content evaluation mechanisms.
• Research Motivation and Related Work:
  • Existing credibility evaluation research highlights the importance of visual cues, tone, and citations, as well as the value of lateral reading.
  • Platforms like Wikipedia and Community Notes demonstrate the positive impact of credibility signals on information evaluation, but similar designs have not been thoroughly explored on video platforms.

Solution

• Method or Solution: The authors propose a prototype system called "Viblio," a Chrome extension that allows users to add and view frame-by-frame citations in YouTube videos, enhancing their ability to evaluate content credibility.
• Innovations:
  • User-Generated Citations: Not only video creators but any user can add citations.
  • Frame-by-Frame Citation Display: Citations are dynamically displayed based on the video timeline.
  • Multi-Category Citations: Provides three categories of citations: supporting the video content, refuting the video content, and offering additional information.
  • Low-Interference Design: To minimize user disruption, the interface aligns with YouTube's design style.
• Implementation Steps and Key Technologies:
  • The Chrome extension was implemented using React and Python Flask technologies, with Azure CosmosDB used to store user-generated citation data.
  • Viblio offers two views: a timeline view and a list view, supporting an interactive frame-by-frame citation display mechanism.

Research Findings

• Specific Findings:
  • Through contextual inquiry, the study identified common credibility evaluation signals used by YouTube users, such as channel name, production quality, and descriptions, but also found that many signals are misunderstood or flawed.
  • Viblio was developed and its intuitiveness and practicality were validated through reflective user studies.
  • Citations were shown to be valuable in helping users confirm suspicious information, especially in videos on sensitive or controversial topics.
• Advantages Compared to Existing Solutions:
  • Addresses the limitations of traditional information panels and existing algorithmic recommendations by providing a more flexible, user-controlled content evaluation tool.
  • Supports lateral reading, enabling users to quickly access external references related to the video for verifying content credibility.
• Experiments or Evaluation Results:
  • Reflective Study:
    • Users generally found Viblio intuitive and easy to use, helping them evaluate video credibility more thoroughly.
    • In science and educational videos, citations were used as tools for extended learning; in controversial topics, citations helped users verify or refute information.
  • Impact Assessment:
    • The average improvement in users' credibility ratings demonstrated the practical impact of citations in enhancing information credibility.
    • Users had mixed opinions on the utility of citations for political/controversial content.
  • Statistics:
    • Participants clicked on citation source articles between 0 and 35 times, reflecting varying levels of engagement with lateral reading.
• Limitations and Future Directions:
  • Granularity and Complexity: For low-traffic videos, citations may be sparse, affecting the system's scalability.
  • Mitigating Malicious Behavior: The current version lacks automated mechanisms to review malicious or low-quality citations.
  • Scalability: Further exploration is needed to apply Viblio to other platforms (e.g., TikTok) and to integrate it into recommendation algorithms.
  • Cognitive Load: Viblio increases cognitive load during video viewing, especially for complex topics, necessitating further design optimization to improve efficiency.

Conclusion

Viblio offers an innovative design that empowers users to better evaluate content credibility, contributing to the prevention of misinformation on social media platforms. Despite facing technical and design challenges, the system provides a new direction for promoting information accuracy and user engagement, with significant potential applications in education, science communication, and controversial topics.