Exposed or Erased: Algorithmic Censorship of Nudity in Art

AI Ethics, Fairness & AccountabilityContent Moderation & Platform GovernanceOnline Identity & Self-PresentationVisual Artists & DesignersContent Governance & Platform Compliance TeamsSociologists & Anthropologists

Title of the Paper

Exposed or Erased: Algorithmic Censorship of Nudity in Art

Paper Information

  • Field of Study: Algorithms and Art
  • Keywords: Content Moderation, Algorithmic Censorship, Artistic Nudity, Online Freedom of Expression, Social Media

Research Background and Issues

  • Identified Problems or Challenges: Artistic nudity is frequently subjected to algorithmic censorship on social media, including content removal and shadow banning. These algorithms exhibit ambiguity in defining and classifying artistic nudity versus pornographic or violent content, often accompanied by technical limitations and societal biases.
  • Significance: Social media has become a crucial platform for modern artists to showcase their work. Algorithmic biases and lack of transparency not only constrain artistic freedom but also have profound impacts on cultural expression and related institutions.
  • Research Motivation and Related Work:
    • Existing studies have primarily focused on the impact of content moderation on general users, with limited attention to artistic content.
    • Moderation rules on social platforms are largely commercially driven, reflecting underlying social and political values, such as the influence of Puritan values from the U.S. on global users.

Proposed Solutions

  • Proposed Approach:

    • Conduct semi-structured interviews with 14 artists to uncover the impact of algorithmic censorship on the creation and sharing of artistic nudity, while exploring potential solutions.
    • Advocate for treating art as a distinct category in moderation policies.
  • Innovative Contributions:

    • Introduce a new perspective on distinguishing artistic nudity from other types of content, emphasizing the importance of considering the creator's "intent" and the context of the work.
    • Highlight the necessity of multi-stakeholder governance (artists, technology developers, policymakers, and the public) and interdisciplinary approaches.
  • Implementation Steps and Techniques:

    • Utilize participatory design to gather artists' perspectives.
    • Propose the development of explainable AI technologies to reduce moderation errors.
    • Advocate for more transparent platform rules and more efficient appeal mechanisms.

Research Outcomes

  • Specific Findings:

    • Documented cases of emotional, financial, and artistic losses caused by algorithmic censorship.
    • Identified significant changes in artists' media habits and creative directions due to censorship, such as self-censorship and shifts in artistic expression trends.
    • Elaborated on the negative impacts of censorship on art institutions and cultural perception, such as the homogenization of mainstream artistic styles.
  • Advantages Compared to Existing Solutions:

    • Provided unique insights into the impact of moderation mechanisms on culture and artistic creation from the perspective of artists.
    • Emphasized a combined approach addressing both technical limitations and governance issues, rather than solely relying on technological improvements.
  • Experimental or Evaluation Results:

    • Identified 55 relevant themes through interview analysis.
    • Synthesized these themes to uncover patterns in censorship behavior and its complexities across technological, cultural, and governance dimensions.
  • Limitations and Future Directions:

    • Limited to the perspectives of Western artists, lacking comprehensive coverage of users from other regions or cultures.
    • Called for strengthening educational systems to promote understanding and acceptance of nude art.
    • Suggested future research to explore multimodal data and human-machine collaboration to improve moderation effectiveness.

Quick Actions

Share

Share this page

ios_share

https://hci.top/en/papers/chi/148324/2024

AdRecommended

Learn AI Coding at CodeNow

open_in_newOpen DOI Link
DOI: https://doi.org/10.1145/3613904.3642586
At a Glance

Paper Snapshot

fact_check
dataset
Source
CHI
calendar_month
Year
2024
emoji_events
Award
No award tagged
group
Authors
3 authors
sell
Subtopics
AI Ethics, Fairness & Accountability, Content Moderation & Platform Governance, Online Identity & Self-Presentation
work
Professions
Visual Artists & Designers, Content Governance & Platform Compliance Teams, Sociologists & Anthropologists
article
Content Status
Full text indexed
hub
Related Papers
0 related papers