Public Opinions About Copyright for AI-Generated Art: The Role of Egocentricity, Competition, and Experience

Research Background and Issues

• Identified Problems or Challenges:
  This study investigates the controversies triggered by breakthroughs in generative artificial intelligence (GenAI) in the fields of art and law, particularly concerning copyright protection for AI-generated artworks. The divergence between legal frameworks and public opinion on these issues, as well as how current laws adapt to this emerging technology, constitutes the primary challenge of the research.

• Importance of the Issue:
  GenAI's ability to generate images has revolutionized the artistic creation process, directly impacting perceptions of creativity and ownership. These controversies not only touch upon philosophical reflections on art and creation but also involve the democratic foundations of legal practice and policy-making.

• Research Motivation:
  While substantial research on AI-generated content has focused on normative discussions, there is a lack of studies addressing public perceptions of these issues. Understanding public attitudes toward the copyright ownership of AI-generated content can both assess the democratic legitimacy of laws and help prevent potential conflicts between legal frameworks and public expectations.

• Related Work:
  Previous studies in the legal domain regarding the protection and ownership of AI-generated content, public awareness of intellectual property, and human perceptions of AI outputs provide the foundational background for this research.

Solution

• Proposed Approach:
  The authors designed an AI art exhibition as an experimental setting to examine public perceptions of creativity, effort, and skill in AI-generated images, and to study participants' tendencies to attribute authorship and copyright under different scenarios.

• Innovative Aspects:
  Few existing studies have systematically investigated how AI-generated content aligns with copyright law from the public's perspective. This experiment combines incentivized art competitions with legal analysis of AI-human collaborative outputs, serving as a practical application of theoretical concepts.

• Implementation Steps and Key Techniques:

  • Design an online AI art exhibition where users can generate and submit artworks using "DALL-E 3."
  • Participants are divided into three roles: creators, related evaluators (who both submit and evaluate artworks), and unrelated evaluators (who only evaluate others' artworks).
  • Submitted artworks are assessed based on multiple copyright-related attributes, including creativity, effort, skill, authorship, and rights attribution.
  • The experimental design tests the effects of egocentric bias, competition, and user experience on public attitudes.

Research Findings

• Specific Results:

  1. Participants identified creativity and effort (but not skill) as key factors required for generating AI art.
  2. Users and contributors to AI model training data were considered the most likely authors of the images, while AI models and development companies received lower attribution.
  3. In scenarios with monetary rewards, AI creators exhibited significant egocentric bias in evaluating their own works, though this bias was less pronounced in purely theoretical copyright issues.

• Comparison with Existing Solutions:
  This study provides additional evidence on public perceptions, complementing research focused on theoretical analysis or legal amendments. The findings highlight the importance of multi-stakeholder distribution models (e.g., users, data contributors).

• Experimental and Evaluation Results:

  • Creators' evaluations were generally higher than those of evaluators for the same artworks.
  • People tended to believe that data contributors should receive more explicit recognition in rights distribution, suggesting that the current model focusing on the interests of development companies may diverge from public expectations.
  • User experience positively influenced attitudes toward generated works, indicating that familiarity with the tool leads to greater recognition of users' dominant roles.

• Limitations and Future Directions:

  • The experiment could not isolate the independent effects of egocentric bias and user experience.
  • Public support for the distribution of monetary rewards between users and data contributors remains unexplored.
  • Legal analysis could be extended to other generative forms, such as music and novels.
  • Future research could design more complex experiments to observe how users' ability to edit AI-generated content affects evaluations of copyright-related factors.

Through this study, the authors emphasize the critical role of public opinion in legal issues related to AI and creation, providing valuable references for future regulations and system design concerning generative AI.