Exploring the Opportunities and Challenges of ChatGPT in Academic Writing: Reply to Bom et al.

Dear Editor,

We would like to address the assumptions made during the roundtable discussion on the use of ChatGPT in academic writing, as reported in the recent editorial [1]. This is a timely topic, since ChatGPT is back after a temporary ban in Italy due to privacy concerns, also having been accused of not verifying user age. While we acknowledge the insights provided by the participants, we believe that there are certain aspects that merit further evaluation and discussion.

Firstly, the discussion seemed to focus primarily on the benefits and limitations of ChatGPT without adequately addressing the potential ethical concerns. ChatGPT and other language models are known to have biases inherent in their training data, which can lead to biased outputs. It is crucial to explore how these biases may impact the scientific integrity of research papers and how they can be mitigated to ensure fair and unbiased representation of knowledge.

We concur with the authors that ChatGPT can be a valuable tool for generating manuscript outlines. As stated by Dr Dong-woo Kim, finding appropriate references for the text generated by ChatGPT is not completely feasible. From what we have experienced, it has a propensity to invent up fake references that are inexistent for the texts generated. Thus, we think that it would be wise to use another search engine to search references for the text produced by ChatGPT.

The discussion on attribution and authorship also requires deeper consideration. While ChatGPT may contribute to the generation of content, it lacks the intent, creativity, and accountability associated with human authors. Recognizing ChatGPT as an author would blur the line between human-generated and AI-generated content, potentially undermining the intellectual contributions and responsibilities of human researchers. Thus, it would be suitable to acknowledge its use in the acknowledgement section of articles. However, it should not be limited to disclosure alone. It should also address the appropriate usage of ChatGPT, emphasizing the importance of critical thinking, independent analysis, and the integration of human expertise alongside AI tools.

Further, in accordance with the ideas expressed by Dr Ki-Seong Park, we think that during writing, it is necessary to request ChatGPT to tailor the answer according to our specific needs. In addition, Dr Jungsu Oh mentioned that he received the same incorrect response from ChatGPT over and over again without an admission of error. In our experience, ChatGPT has a tendency to approve our statements as correct. For example, if we asked it “are you sure?”, it will reply “I apologize for my previous answer” and will most likely mention another answer even when the initial answer was correct.

Undoubtedly, one evident advantage is the accelerated pace of research and publication facilitated by AI tools. This allows researchers to allocate more time to the actual study rather than the writing process. However, it is important to consider the current academic landscape, where tenure is often influenced by metrics such as the H index, which is based on the number of citations an author receives from other authors. If AI tools were to be used by scholars solely for the sole purpose of boosting publication output, it could potentially overwhelm journal editors and lead to an overabundance of AI-generated content in important databases like PubMed and Scopus. This content would be primarily created by AI with minimal human input. The repercussions of this trend on future research remain uncertain.

One fact that was not addressed during the round table discussion was regarding LLM and ChatGPT detectors. Our opinion is that ChatGPT detectors may not be fully reliable to date in discriminating between human-generated text and AI-generated content and thus may cause inconveniences to authors during editorial processing of manuscripts.

Also, in Italy, as per legal requirements, medical students must defend a thesis before obtaining the title of Doctor of Medicine (MD). I agree with you that ChatGPT may be useful in writing a thesis. However, as the utilization of LLM technologies becomes increasingly prevalent in the future, it is necessary to reassess the purpose and significance of writing a dissertation, considering that a substantial portion of the work may be accomplished using these advanced technologies.

In conclusion, while the roundtable discussion shed light on some aspects of using ChatGPT in academic writing, it is imperative to critically examine the implications of AI tools like ChatGPT and develop comprehensive guidelines that uphold academic integrity, transparency, and responsible use.

Author Contribution

Arosh S. Perera Molligoda Arachchige participated in conceptualization and writing—review and editing. Niccolò Stomeo participated in writing—original draft.

Data Availability

Not applicable.

Declarations

Conflict of Interest

Arosh S. Perera Molligoda Arachchige and Niccolò Stomeo declare no conflicts of interest.

Ethical Statement

Not applicable.

Consent to Participate

Not applicable.

Consent for Publication

Not applicable.