This document proposes a framework called PaperCard for transparently documenting machine assistance in academic writing. With advances in generative AI, machines can now play a larger role in the writing process. However, this raises legal and ethical issues regarding fairness, accountability, and transparency that have yet to be addressed. PaperCard would require authors to declare any machine assistance, including details about the specific AI system, prompts provided, and edits made to generated content. This framework aims to govern the use of machines in academic writing by promoting transparency around their involvement.
Artificial intelligence in the post-deep learning era
Reporting Machine Assistance in Academic Papers
1. Partner/Sponsor:
PaperCard for Reporting Machine Assistance in Academic Writing
WON IK CHO∗1, EUNJUNG CHO∗2, KYUNGHYUN CHO3
1Seoul National University, South Korea; 2 ETH Zürich, Switzerland; 3New York University and Genentech, USA
• With rapid developments in generative AI, machines can
now play a more major role in our writing process
• However, this also brings about various legal and ethical
issues that are yet to be addressed
• We discuss and address these issues by proposing
PaperCard, a documentation framework for authors to
transparently declare the use of machines in their writing
process
1. How have technological developments changed the
academic writing process? How have recent developments
in generative AI influenced the concept of authorship?
2. What are some key concerns surrounding fairness,
accountability, and transparency related to the use of
machine assistance in academic writing (especially against
the backdrop of the recent developments in generative AI)?
3. How should we govern the use of machines in academic
writing?
This manuscript was written with machine assistance: Yes
Depth of Assistance We used OpenAI ChatGPT to generate
some 'original' content for the following sections: [section
names]. We gave ChatGPT detailed outlines of each section as
input prompts. All prompts were drafted by us. We did some
heavy editing of the generated content, by adding more depth
and insight through additional research, for which we did not use
ChatGPT. The remaining sections were entirely written by us.
Declaration As per OpenAI's terms of use, authors own the
right of the generated text and are accountable for potential
conflicts. We believe the AI-generated texts included in this
paper do not have elements that may give rise to ethical issues.
We also inspected the texts thoroughly to check for their
academic accuracy and plagiarism.
Details We adopted ChatGPT Version Jan. 9, 2023 provided by
OpenAI, accessed 2023.01.11 - 18. We created a set of prompts
to generate content for the following sections: [section names].
Summary and details of the prompts are available in Appendix.
Also, we checked the reproducibility of our prompting with
ChatGPT Version Jan. 30, 2023, provided by OpenAI, accessed
in 2023.02.02. Results are available in Appendix.
• Advent of typewriter, internet, Google translator,
Grammarly, QuillBot, etc. streamlined academic writing
process, but had relatively limited impact on authorship
• Recent developments in generative AI pose new
challenges. Many journals / conferences introduced ban or
extra requirements on using generative AI in academic papers
FAIRNESS
• Accessibility: Not everyone can afford / access the AI tools
• Low-resource languages: Performance differences exist
• Copyright and licensing: Who should be the beneficiaries?
5. PaperCard for Practitioners (Sample)
4. Ethical and Legal Issues
3. Technology, Generative AI, and Authorship
References
1. Timnit Gebru, Jamie Morgenstern, Briana Vecchione, Jennifer Wortman Vaughan, Hanna Wallach, Hal
Daumé Iii, and Kate Crawford. 2021. Datasheets for datasets. Commun. ACM 64, 12 (2021), 6–92.
2. Margaret Mitchell, Simone Wu, Andrew Zaldivar, Parker Barnes, Lucy Vasserman, Ben Hutchinson, Elena
Spitzer, Inioluwa Deborah Raji, and Timnit Gebru. 2019. Model cards for model reporting. In Proceedings of
the conference on fairness, accountability, and transparency. 220–229
3. OpenAI. 2022. Terms of use. https://openai.com/terms/
4. ACM. 2023. ACM policy on Authorship. https://www.acm.org/publications/policies/authorship-policy
5. Nature. 2023. https://www.nature.com/nature-portfolio/editorial-policies/authorship
6. Program Chairs. 2023. ACL 2023 policy on Ai Writing Assistance. https://2023.aclweb.org/blog/ACL-2023-
policy/
7. ICML. 2023. https://icml.cc/Conferences/2023/llm-policy
1. Summary and Contributions
2. Research Questions
3. Authorship in Academic Writing
Intellectual
contribution
Participation in
writing
Approval of
final manuscript
• Two main questions arise:
1) Should we merit authorship to AI?
2) Does using AI undermine contribution of human
authors? How should we measure / evaluate the
contribution of human authors vs. AI?
No concrete conclusions can be drawn at this point
ACCOUNTABILITY
• Who should be held accountable for the generated content?
TRANSPARENCY
• Challenges in accurate evaluation of human contribution
Editor's Notes
Technological advancements in computing, especially large language models such as GPT variants, have streamlined the academic writing process with exceptional capabilities in text generation and user alignment
The increasing prevalence of AI-generated text in academic writing necessitates makes it difficult to distinguish human and machine contributions, and necessitates clear guidelines for its responsible use in research papers and mitigating ethical and legal issues regarding authorship, licensing, accountability, and potential copyright infringements