See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/368857296

Concerns with the Usage of ChatGPT in Academia and Medicine: A Viewpoint

Article in American Journal of Medicine Open · February 2023

DOI: 10.1016/j.ajmo.2023.100036

CITATIONS READS
16 870

2 authors:

Muath Alser Ethan Waisberg

Cairo University University of Cambridge
18 PUBLICATIONS 694 CITATIONS 58 PUBLICATIONS 228 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Clinical Audit View project

Gaza medical students abroad: motives, siege and wars effect, academic performance and career prospects View project

All content following this page was uploaded by Muath Alser on 25 May 2023.

The user has requested enhancement of the downloaded file.

 American Journal of Medicine Open 9 (2023) 100036
Contents lists available at ScienceDirect
American Journal of Medicine Open
journal homepage: www.elsevier.com/locate/ajmo
Concerns with the Usage of ChatGPT in Academia and Medicine: A
Viewpoint
We have written this article to express our concerns about the grow-
ing trends of ChatGPT usage in academia and medicine. ChatGPT is an
artificial intelligence large language model developed by OpenAI and
launched on 30th November 2022 as a free beta version in many coun-
tries worldwide. This language model can produce human-like text re-
sponses to users’ prompts in almost any field.
ChatGPT Authorship
Recently, 2 articles have been published in peer-reviewed medical
journals and 2 other preprints where ChatGPT was credited author-
ship.1-4
As per the International Committee of Medical Journal Editors
(ICMJE) guidelines for authorship, to be eligible for authorship, you
need to fulfill all these 4 criteria5 :
1. “Substantial contributions to the conception or design of the work;
or the acquisition, analysis, or interpretation of data for the work;
AND
2. Drafting the work or revising it critically for important intellectual
content; AND
3. Final approval of the version to be published; AND
4. Agreement to be accountable for all aspects of the work in ensuring
that questions related to the accuracy or integrity of any part of the
work are appropriately investigated and resolved.”5
In the case of ChatGPT authorship, in 3 out of the 4 examined papers,
ChatGPT made a substantial contribution in writing parts of the paper
(only 1 criterion applies),1, 2, 4 while in the remaining paper, ChatGPT’s
answers were further analyzed by human authors, so ChatGPT served
as a research participant, not an author (no criteria applies).3 In all of
these papers, ChatGPT did not approve the final version of the published
article, nor did it agree to be responsible for the publication’s accuracy
or declare any conflict of interest. And even if the chatbot approved
these publications, should we consider its approval legally significant?
ChatGPT Plagiarism
We performed a plagiarism check on the parts of writing contributed
by ChatGPT in 1 of these articles using Grammarly, Plagiarism Checker
X, and PlagScan (by Turnitin). Unsurprisingly, plagiarism was detected
after manual review, ranging from 5% to 33.8% and reaching 48.9%
after merging all of the scans (see supplementary file).
https://doi.org/10.1016/j.ajmo.2023.100036
Received 15 February 2023; Accepted 21 February 2023
Available online 27 February 2023
2667-0364/© 2023 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)
Some phrases detected by these tools were copied word for word
by ChatGPT from unreliable sources like Wikipedia, LinkedIn, Medium,
and even the Apple App Store (direct plagiarism).
ChatGPT did not provide by default any citations for its output (para-
phrasing plagiarism). When asked for references for some of its answers,
it provided citations that do not exist (source-based plagiarism). When
asking ChatGPT the same prompt again, it gives very similar answers,
which could lead to self-plagiarism.
Including ChatGPT’s output as it is in the manuscript without quo-
tation marks gives the impression that this part is contributed by all
authors (misleading attribution). Citing ChatGPT for its output is uneth-
ical, too, as the primary source of information is not cited (source-based
plagiarism). Hence plagiarism is inevitable when using ChatGPT.
ChatGPT Sources of Bias
Up to this point, OpenAI has not published the source code for Chat-
GPT, nor has it explained its learning data sources. Asking ChatGPT
about its data updates, it mentioned that its knowledge is limited to
2021, which means that it is outdated for 2023.
Based on our plagiarism checks, ChatGPT uses academic and nonaca-
demic sources and apparently does not differentiate between sources of
information based on their level of evidence. This accounts for the errors
that ChatGPT can occasionally make.
Developers have fine-tuned the output of ChatGPT through super-
vised and reinforcement learning (reward system for desired answers);
this could bias the output toward the developers’ opinions. More seri-
ously, public users can participate in this tuning through upvoting or
downvoting answers; this manipulation devalues the output from a sci-
entific point of view.
ChatGPT in Medical Education and Clinical Practice
One of the aforementioned preprints that gave authorship to Chat-
GPT suggested the use of language models in medical education and in
leading clinical decisions as the chatbot passed a USMLE exam!3 Many of
the USMLE exams are publicly available online with answers. The pass-
ing of USMLE by this bot reflects its ability as a search engine rather
than as a reliable source for clinical decisions.
This hypothesis is supported by a Korean study that compared Chat-
GPT’s performance in a parasitology exam with medical students’ perfor-
mance and found that ChatGPT scored significantly less than the medical
students.6
 American Journal of Medicine Open 9 (2023) 100036

Final Remarks Muath Alser∗

Cairo University Faculty of Medicine, Cairo, Egypt
ChatGPT is a language model that analyzes statistical patterns of
Ethan Waisberg
language used throughout a large data set. This makes this bot capable
University College Dublin School of Medicine, Belfield, Dublin, Ireland
of producing appealing writings language-wise, but it lacks depth and
factual accuracy. Due to the previously explained serious drawbacks
of this ChatGPT version, we should recommend against its usage in ∗ Corresponding author: Muath Alser, Cairo University Faculty of
academia. In case of the unavoidable need for the use of this bot in scien- Medicine, Al-Saray Street, El Manial, Cairo 11956 Egypt.
tific publications, we should acknowledge the bot, but not grant author- E-mail address: moathalser@gmail.com (M. Alser)
ship, while paying attention to different types of plagiarism and biases
it has. References

1. O’Connor S. ChatGpt. Open artificial intelligence platforms in nursing education: tools

Declaration of Competing Interest for academic progress or abuse?. [published Online First: 2022/12/23]. Nurse Educ
Pract.. 2023;66:103537. doi:10.1016/j.nepr.2022.103537.
2. Zhavoronkov A. Rapamycin in the context of Pascal’s Wager: generative pre-
The authors declare that they have no conflict of interest. trained transformer perspective. Oncoscience. 2022;9:82–84 [published Online First:
2023/01/03]. doi:10.18632/oncoscience.571.
3. Kung TH, Cheatham M, Medenilla A, et al. Performance of ChatGPT on USMLE: po-
Funding tential for AI-assisted medical education using large language models. medRxiv. 2022
2022.12.19.22283643. doi:10.1101/2022.12.19.22283643.
4. Gpt Generative Pretrained Transformer, Thunström A, Osmanovic, Steingrimsson S.
This research did not receive any specific grant from funding agen- Can GPT-3 write an academic paper on itself, with minimal human input? 2022. Avail-
cies in the public, commercial, or not-for-profit sectors. able at: https://hal.science/hal-03701250.
5. ICMJE. Recommendations for conduct, reporting, editing, and publication of scholarly
work in medical journals: defining the role of authors and contributors. Available at:
Supplementary materials https://www.icmje.org/recommendations/browse/roles-and-responsibilities/defining-
the-role-of-authors-and-contributors.html. Accessed January 20, 2023.
6. Huh S. Are ChatGPT’s knowledge and interpretation ability comparable to those
Supplementary material associated with this article can be found, in of medical students in Korea for taking a parasitology examination?: a descriptive
the online version, at doi:10.1016/j.ajmo.2023.100036. study. [published Online First: 2023/01/12]. J Educ Eval Health Prof.. 2023;20:1.
doi:10.3352/jeehp.2023.20.01.

View publication stats