I. Introduction
The use of Large Language Models (LLMs), such as ChatGPT, in academic writing, offers researchers a wide range of potential benefits, including the ability to generate large volumes of text quickly and efficiently. However, this emerging technology also presents a significant challenge to the transparency and credibility of academic writing. Therefore, to ensure the responsible and transparent use of LLMs, JTS team has developed the following guidelines for our authors.
II. Guidelines on Transparency and Credibility Guide to LLM use:
1. LLM tools will not be accepted as credited authors. This is because any attribution of authorship carries with it responsibility for the work, and AI tools cannot assume such responsibility.
2. Authors who utilize LLMs in their research must explicitly disclose this fact in their manuscript, including the name and version number of the model used. This disclosure should be provided in the methodology section or acknowledgments section of the manuscript.
3. Authors must describe in detail the method or acknowledgments section they used to generate text with the LLM, including the parameters and settings they used. Any additional pre-processing or post-processing steps applied to the generated text must also be reported.
4. Authors must ensure that the generated text is factually accurate, relevant to the research question, and consistent with the existing knowledge in the field.
5. Authors must clearly distinguish between the original writing and the text generated by the LLM. The generated text should be italicized and enclosed in quotation marks.
6. Authors must acknowledge the potential limitations of LLMs in their manuscript, including the potential for bias and errors. Authors should also explain how they addressed these issues to the best of their ability.
7. The use of LLMs in research must be justified based on the specific research question and the potential benefits of using the tool. The manuscript should explain the reasons for using LLMs and how it contributed to the research question.
8. The editorial team at JTS reserves the right to request access to the raw data and code used in LLM-assisted research to verify the authenticity of the results.
III. The responsible use of ChatGPT in academic writing
Consider the following examples:
First example generated by ChatGPT v2.0:
"We used ChatGPT v2.0 to generate text for our literature review. We set the length of the generated text to 500 words and used a top-k sampling strategy with a k value of 40. We then manually reviewed the generated text to ensure its accuracy and relevance to our research question. Any irrelevant or inaccurate text was removed. We acknowledge that ChatGPT may have inherent biases, such as the overrepresentation of certain topics, and we attempted to address this by incorporating multiple sources of information and conducting a manual review. The generated text is clearly distinguished from our original writing by italicizing and enclosing it in quotation marks. The use of ChatGPT was justified by its ability to quickly generate a large volume of relevant literature, which facilitated the efficiency of our literature review process."
Second example:
“Declaration of AI-usage
For drafting parts of the presented text, we used the pre-trained ChatGPT (3.5, 2023-Jan-9) model developed by OpenAI as described in the paper by Brown et al. (2020). We would like to acknowledge OpenAI for providing the pre-trained model and the authors of the paper for their work on GPT models. All prompts and their generated texts are available in the Supplement. We checked the text for unintended plagiarism, and we verified all facts and references that we used from LLM outputs.” (Aczel, Balazs, and Eric-Jan Wagenmakers. 2023. “Transparency Guidance for Chatgpt Usage in Scientific Writing.” PsyArXiv. February 6. doi:10.31234/osf.io/b58ex.)
IV. Final considerations
By adhering to these guidelines, JTS aims to maintain the integrity and transparency of academic writing and ensure that all contributions are appropriately attributed. The use of LLMs in research presents exciting opportunities for advancing knowledge creation, and we encourage the responsible use of these tools by our authors.
V. AI Usage Statement
To draft portions of this policy, the editors used the pre-trained ChatGPT Plus (3.5, 2023-Feb-23) template developed by OpenAI to assist in proofreading the texts, drafting the example, and clarifying GPT authorship in academic writings. We would like to thank OpenAI for providing the ChatGPT template. The data was obtained from the paid version of ChatGPT Plus subscribed by Telos Publicações e Serviços Ltda.