"It Matches My Worldview": Examining Perceptions and Attitudes Around Fake Videos

Title of the Paper

"It Matches My Worldview": Examining Perceptions and Attitudes Around Fake Videos

Paper Information

• Domain: Research on the dissemination of fake videos on social media and user behavior
• Keywords: Fake videos, deepfakes, misinformation, India, Global South, ICTD (Information and Communication Technology for Development), HCI4D (Human-Computer Interaction for Development)

Research Background and Issues

• Issues and Challenges:

  • With the increasing adoption of social media among users in the Global South, fake videos (including both conventional fake videos and AI-generated deepfake videos) have become significant channels for spreading misinformation.
  • In a low-resource environment like India, where internet users often have limited digital literacy and tend to trust video content, identifying fake videos poses a greater challenge.
  • Existing research predominantly focuses on user behavior in the Global North (e.g., the United States) and has not adequately explored how users in the Global South perceive, encounter, and handle fake videos.

• Significance of the Study:

  • Fake videos not only threaten personal privacy and security but can also exacerbate social divisions and political polarization, particularly in resource-constrained regions.
  • Given India's vast number of internet users, gaining a deeper understanding of their motivations and behaviors can help design more effective mechanisms to curb the spread of fake videos.

• Research Motivation:

  • To explore the perceptions, threat awareness, and sharing behaviors of diverse social media users in rural and urban India regarding fake videos.
  • To address the academic gap in understanding how users in the Global South respond to and act upon fake videos.

Solutions

• Research Methods:

  • Conducted semi-structured interviews with 36 social media users from rural and urban India, who varied in literacy levels, digital skills, and social media experience.
  • Used video probes, including one real video and three deepfake videos, to observe participants' ability to identify fake videos and their responses.

• Key Innovations:

  • Focused on the high-risk context of India, employing video probes to guide interviews and uncover psychological and threat models related to various types of fake videos.
  • Provided integrated design recommendations to help social platforms reduce the spread of fake videos.

• Key Techniques and Steps:

  1. Study Design and Data Collection:
     • Recruited participants from rural and urban areas and sent video probes via WhatsApp.
     • Conducted online interviews to ensure participant safety.
  2. Data Analysis:
     • Translated and transcribed interview audio recordings, using inductive thematic analysis to extract core themes.
  3. Methodological Rigor:
     • Employed multiple rounds of coding and peer cross-checking to ensure the reliability of data analysis, and conducted member checks to validate conclusions.
  4. Design Recommendations:
     • Proposed specific, culturally sensitive solutions to address issues of low digital literacy, lack of knowledge, and insufficient tools among users.

Research Findings

• Key Findings:

  • Knowledge and Cognitive Differences: Most users had no awareness of deepfake videos, with only a few recognizing that videos could be edited or segmented.
  • Behavioral Patterns:
    • Most users ignored potentially fake videos, while a few deleted or blocked such content.
    • Some users believed that fake videos aligning with their worldview were worth sharing, even if they knew the videos were inaccurate.
  • Impact Assessment:
    • Fake videos were described as more harmful than text or image-based rumors, particularly in their significant impact on public discourse and social stability.
    • Female participants specifically highlighted the threats posed by deepfake pornographic videos to personal privacy and psychological safety.

• Comparison with Existing Solutions and Advantages:

  • This study is among the few qualitative studies focusing on users in the Global South, addressing gaps in regional and cultural specificity in existing research.
  • Proposed a range of specific measures, such as credibility labels powered by explainable AI, reflective sharing mechanisms, and regular digital literacy training.

• Experimental or Evaluation Results:

  • Interviews revealed that rural users often trusted videos received from family and friends, while urban users demonstrated slightly better skills in identifying fake videos.
  • The study uncovered widespread dissatisfaction with current platform reporting features, citing lack of transparency and unclear feedback mechanisms.

• Limitations and Future Directions:

  • Limitations:
    • Data collection relied on interviews, which may have led participants to avoid discussing sensitive topics.
    • The sample was limited to the Indian context, necessitating validation across more cultural and geographic settings.
  • Future Directions:
    • Develop training tools for identifying fake videos tailored to users with low literacy and digital skills.
    • Explore mixed-method approaches to further validate user behavior patterns through quantitative research.
    • Design and test culturally relevant, explainable AI tools and user interfaces to curb misinformation.