Over the past few months, the Internet has been abuzz with talk about ChatGPT, which has put a spotlight on the increasingly advanced capabilities of artificial intelligence (AI) writing tools. Built by OpenAI and based on generative pre-trained transformer (GPT) architecture, ChatGPT is a large language model (LLM) chatbot that has an uncanny ability for human-like responses that seem surprisingly accurate. It’s ability to quickly generate a range of written content such as articles, poems, lecture notes and even software codes has made ChatGPT the talk of the town. Now, Google has come up with its own AI chatbot called Bard AI, adding to the already heightened excitement surrounding AI writing tools and their impact on the world.
ChatGPT makes its debut as author on research papers
ChatGPT’s content generating abilities are not only creating waves across businesses and industries, they have also percolated down into the world of science. The AI writing tool is being used in several ways including writing essays and creating research paper abstracts. In fact in a recent study, academic reviewers only detected 63% of AI-generated abstracts created by ChatGPT.1 Researchers have also published scholarly articles and preprints that have assigned ChatGPT with authorship in research. This has raised an important question – is it appropriate to cite a bot as an author in research?
Authorship in research isn’t the same as authorship of a newspaper article or a book. Authors of newspapers and books are commissioned and paid to write, unlike researchers who depend on successful grant applications. The reward for researchers is linked to authoring research papers that are cited and have an impact in their respective fields; this in turn helps them grow their academic career and build their reputation as credible researchers.
Moreover, authorship in research papers doesn’t need you to have actually “written” the research paper, rather, being credited as an author is a reflection of your genuine and significant contribution and involvement in the research study. Authorship in research carries with it a responsibility for the quality and accuracy of the contents of the research paper. This is clearly mentioned in the 2018 Australian Code for the Responsible Conduct of Research, which states
“All listed authors are collectively accountable for the whole research output. An individual author is directly responsible for the accuracy and integrity of their contribution to the output.” 2
Using AI writing tools like ChatGPT raises concerns about the validity of the text they generate, and how they can be held responsible as ‘authors’. This has emerged as a particularly important question in subject areas such as medicine, where errors in the research content have the potential to impact people’s lives. In simple terms, if ChatGPT’s contribution to a research paper backfires, who would be held accountable?
Publishers unwilling to accept ChatGPT as an author
While one cannot deny the advantages that ChatGPT brings to researchers, increasing instances of crediting the AI writing tool with authorship in research has pushed top publishers to act. Editors-in-chiefs of two leading academic publishers, Science and Springer Nature, have already declared their positions on assigning ChatGPT authorship in research and using it to generate research articles.
According to Springer Nature, while ChatGPT did not meet the basic requirement for authorship in research, they had no problems with researchers using the AI writing tool as long as its contribution was properly disclosed.3 “An attribution of authorship carries with it accountability for the work, which cannot be effectively applied to LLMs,” says Magdalena Skipper, editor-in-chief of Nature in London.3 Authors using LLMs while developing a paper may document their use in the methods or acknowledgements sections, if required, she added.
While Science journals shared this view on authorship in research, the publishers are clear about not accepting AI-generated content. Holden Thorp, editor-in-chief of the Science family of journals, says, “We would not allow AI to be listed as an author on a paper we published, and use of AI-generated text without proper citation could be considered plagiarism.” 3 Science has also updated its editorial policies to explicitly state that use of AI-generated text (including that generated by ChatGPT), graphics, images, or figures will be considered a serious violation.4
AI writing tools for researchers: What does the future look like?
There are several AI writing tools already available, and with new entrants like Bard AI, this space is bound to see an upward trend. AI tools like ChatGPT for research writing have become a reality. While global publishers are unlikely to allow ChatGPT and other such tools to be credited with authorship in research, there’s nothing stopping researchers from using these tools to strengthen their work in other ways. With the existing uncertainty about authorship in research, the arrival of ChatGPT and the frenzy around it is a wake-up call for the academic community to take a closer look at making things better.
References
- Else, H. Abstracts written by ChatGPT fool scientists. Nature, News, January 2023.https://www.nature.com/articles/d41586-023-00056-7
- Australian Code for the Responsible Conduct of Research, 2018. NHMRC. Available at https://www.nhmrc.gov.au/about-us/publications/australian-code-responsible-conduct-research-2018
- Stokel-Walker, C. ChatGPT listed as author on research papers: many scientists disapprove. Nature, January 18, 2023. https://www.nature.com/articles/d41586-023-00107-z#ref-CR1
- Thorp, H. ChatGPT is fun, but not an author. Authors Info & Affiliations, Science. January 26, 2023. https://www.science.org/doi/10.1126/science.adg7879
