Recent advances in large language models (LLMs) and generative artificial intelligence (AI) have led to the rapid integration of AI tools into research. These tools are now used to assist with literature reviews, data management and analysis, data synthesis, and editing scientific writing. Additionally, many medical journals utilise AI for tasks such as plagiarism detection, image manipulation identification, and identifying expert peer reviewers, helping to streamline the publication process.


The rapid integration of AI tools into academic publishing presents challenges that both authors and editors must address together to avoid undermining the goal of curating and disseminating new knowledge. Issues include LLMs producing outdated or factually inaccurate content, biases in algorithm output, and lack of transparency in how these tools operate. Since peer review depends on human judgment, the unregulated use of AI tools could compromise the quality and trust in the publishing process. While AI tools offer significant benefits, it’s important to balance enthusiasm with caution regarding their limitations and flaws.

 

The editors of the American Thoracic Society (ATS) journals prioritise ensuring the accuracy of published manuscripts and protecting researchers’ work during the confidential peer review process. With the increasing use of AI tools in publishing, there is increased scrutiny of editorial policies regarding AI. ATS editorial leadership emphasises the need to review and discuss AI-related policies, recognising the uncertainties and dynamic nature of AI developments. Ongoing policy revisions will be necessary, requiring transparency and collaboration with authors, reviewers, readers, and other journals to create a suitable framework for AI in academic publishing.

 

ATS policies on the conduct, reporting, editing, and publication of scholarly work, including AI-assisted technologies, follow best practices from leading organisations in scientific publishing. However, due to the rapid evolution of AI technologies, there are still many areas of uncertainty for authors, reviewers, and editors that are not fully addressed by current policies.

 

These policies for authors align with other publishers by prohibiting LLMs from being credited as manuscript authors. Authors must transparently document any use of AI tools in their work, including describing AI tool use in the methods, acknowledgment, or other relevant sections. At submission, authors are required to declare if an AI tool was used, specify the tool, and explain how it was applied in the research. This includes AI tools used for image creation or editing, which must comply with ATS guidelines. The policy ensures transparency and clarity for reviewers and readers, with all manuscripts now including an AI declaration, similar to conflict-of-interest statements.

 

ATS journal leaders emphasise the importance of protecting authors’ intellectual and confidential property during the review process. As a result, the journals prohibit peer reviewers from using LLMs to prepare reviews, as this could involve uploading manuscript content to an AI tool outside the journal’s editorial control. 

 

The policy prohibiting AI-generated peer review comments could discourage diversity among reviewers, particularly for those whose first language is not English. To address this, the revised policy allows reviewers to use LLMs to assist in formulating the language of their comments, as long as manuscript content is not shared with the AI tool. Reviewers are still responsible for the content of their reviews and must declare the extent of their use of LLMs when submitting their review.

 

The leaders of the ATS journals believe AI should assist, not replace, human expertise. The academic community must embrace AI responsibly, avoiding both indiscriminate use and complete avoidance. As the field evolves, policies will be updated, but they will remain focused on the principles that authors are responsible for the integrity of their work and that transparent, full declarations of AI tool use by authors and reviewers are essential.

 

Source: AJRCCM
Image Credit: iStock 

 


References:




Latest Articles

AI, Artifical Intelligence, The American Thoracic Society, scientific research, Large Language Models, LLMs, academic publishing Role of Artificial Intelligence in Scientific Research and Review