Generative AI and Authorship

Presented by Robin Kear, Academic Librarian at the University of Pittsburgh

Introduction

Robin Kear discusses the question: Can generative AI (GenAI) be an author? She explores the implications of this question, considering the rapid advancement of AI technology and its impact on authorship, creativity, and responsibility.

Can GenAI Be an Author?

Kear reflects on her concerns regarding AI's potential to become sentient or possess its own consciousness and agency. She believes that, with the current structure of generative tools, the answer is no. GenAI reacts, suggests, anticipates, and amalgamates existing content but does not create something entirely new.

AI-Generated Content and Authorship

Using an example of an image created by a human using DALL-E (an AI image generator), Kear prompts the audience to consider where authorship resides in such creations. She emphasizes the importance of understanding the human aspects of being an author and creator.

What Makes an Author?

Kear identifies four key human aspects of authorship:

Creativity: The idea must originate from the individual. While influenced by experiences and environments, humans create new things that didn't exist before.
Agency: Authors have the will to decide what to do with their ideas, choosing how, when, and what to produce.
Moral Responsibility: Authors are morally accountable for what they put into the world, and their work should be discoverable and attributable to them.
Legal Responsibility: Authors accept legal responsibility for their creations in the public and economic spheres, including the publishing industry.

Research on AI and Authorship in Academic Journals

Kear shares a research project conducted with colleague Amy Jenkins, examining how research journals are addressing AI and authorship. They analyzed top journals across various disciplines to find policies and guidance on AI authorship.

Methodology

Used Journal Citation Reports to identify impactful journals.
Selected top three journals in chosen categories based on impact factor.
Searched journal and publisher websites for AI authorship policies.

Findings Based on the Four Aspects of Authorship

Creativity and Agency

AI Cannot Be an Author: All journals agreed that an author must be a human being.
Lack of Agency: AI does not have the ability to act independently or be accountable.
AI in Images: Generally not permissible, especially in scientific contexts due to potential harm to scientific advancement.
Writing Assistant vs. Data Analysis: A nuanced difference exists between using AI as a writing tool and using it for data insights, which requires disclosure.

Moral Responsibility

Personal Accountability: Authors must be accountable for their content, hence AI cannot be an author.
Disclosure Requirement: Use of AI tools must be disclosed, with specifics on how and where it was used.
Publication Process: Different guidelines exist for authors, peer reviewers, and manuscript reviewers.
Confidentiality Concerns: Public AI tools like ChatGPT should not be used for peer review due to confidentiality and proprietary rights.

Legal Responsibility

Liability: Journals could be held liable for AI-generated content, so responsibility is shifted to the author.
Verification: Authors are responsible for verifying the accuracy of AI-generated content, including potential errors or plagiarism.
Ethical Breaches: Authors are liable for any breaches of publication ethics, even if AI tools were used.
Guidance from COPE: The Committee on Publication Ethics emphasizes authors' full responsibility for their manuscripts.

Reconsidering the Role of AI in Creative Endeavors

Kear poses critical questions about how we should view AI in the context of creativity:

Should AI be considered an assistant or helper rather than a creator?
Can AI serve as a sounding board for ideas or help augment human creativity?
Where is the ethical line between presenting something as one's own idea versus a technology-created idea?
Given that AI responses are derivative, what is its usefulness in creative work?

Reflection on Automated Creativity

She references the 1982 World's Fair painting robot as an early example of automated creativity, noting that while simplistic compared to current AI, it prompts consideration of the evolving role of technology in authorship.

Further Considerations

Kear discusses additional points stemming from her findings and university discussions:

Changing Acceptance: The use of AI in writing may become more accepted over time, potentially becoming seamless and expected.
Reflecting Existing Challenges: AI often mirrors societal biases and existing challenges related to transparency, integrity, and accountability.
Core Principles: The fundamental principles of research and publishing should continue to guide the use of AI in authorship.

Question and Answer Session

To What Extent Do Humans Also Derive from Other Content?

Response: Kear acknowledges that humans are influenced by their environment and existing works. In academic writing, literature reviews are essential for building upon previous research, but authors strive to contribute something new to the conversation.

At What Point Is AI Used or Not Used?

Response: She differentiates between general writing tools (like Microsoft Editor or Grammarly) and generative AI tools. While tools like Microsoft Co-Pilot are still developing, she focuses on the implications of generative AI in authorship.

If a Student Uses an AI Tool to Fully Write a Paper, Who Is the Author?

Response: Kear advises against students using AI to write entire papers. Such papers may contain inaccuracies, lack depth, and could be easily identified by instructors. Students should be cautious about relying on AI for academic work.

Future Value of Writing in Editing vs. Writing Itself

Response: Currently, the value of generative AI lies in its ability to assist rather than replace human creativity. She mentions authors using AI tools based on their own work to aid in writing, but emphasizes that AI should complement, not replace, human authorship.

Conclusion

Kear concludes by emphasizing the importance of maintaining core principles in research and publishing as AI continues to evolve. Transparency, integrity, attribution, and accountability should guide any use of AI in authorship and creative endeavors.