ARTIFICIAL INTELLIGENCE AND SCIENTIFIC PRODUCTION: OPPORTUNITIES AND CHALLENGESArtificial Intelligence (AI) is a constantly evolving field, focused on developing machines and systems that mimic aspects of human intelligence to offer solutions to a wide range of problems and challenges. Although initially popularized by science fiction in the early 20th century, the idea of machines capable of simulating human behavior and "thinking" was proposed by the British mathematician Alan Turing, widely recognized as the "father of computer science." However, the term "Artificial Intelligence" was formally introduced by John McCarthy, in 1956, during the Dartmouth Summer Research Project on Artificial Intelligence conference at Dartmouth University in the United States, the initial milestone of academic and scientific discussions on the subject1-2.
Historically, AI derives from human curiosity or from an invention that has been developing since the 1960s and 1970s. However, computer technological development and internet networks at that time did not have the capacity to create algorithms that form the neural networks we know today. Thus, AI has moved beyond being a fascinating narrative of human aspirations, technological innovations, and major scientific discoveries, and is increasingly becoming a part of everyday activities that influence sociality and impact decisions once exclusively made by humans3.
As AI transitions from fiction, we are living in the era of "big data," where the ability to collect and process large volumes of information becomes increasingly accessible, and its application proves promising in various sectors of society. Furthermore, AI is already integrated into our daily lives through personal assistants (such as Siri, Alexa, and Google Assistant), navigation apps (Waze and Uber), aviation, virtual assistants (chatbots), and wearable devices such as smart clothing and accessories (smartwatches), among others1.
In 2023, AI applied to services and solutions experienced exponential growth, and its effectiveness in various areas got widespread recognition, including its relevance as tools in writing and scientific research4. However, despite significant advances, criticisms, ethical concerns, and disagreements persist regarding the limits and implications of using these technologies, especially in the academic environment. This ongoing debate highlights the need to reflect on the opportunities and challenges that accompany the growing application of these innovations in scientific production.
Among the AI tools that can be used in scientific production, ChatGPT® emerges as one of the most popular today. This chatbot enables productive dialogue by offering valuable insights and facilitating brainstorming sessions to refine research questions. Additionally, it allows for the refinement of parts of the text, whether it's an idea, context, or intended outcome, providing additional opinions. The quality of the prompt, the set of words used for the language model to generate AI content, directly influences the quality of the response. When it comes to improving writing in English, for example, ChatGPT® stands out by editing texts, correcting grammatical errors, punctuation, and spelling, as well as enriching vocabulary. Additionally, it can be employed to reduce text length and properly format references. In summary, ChatGPT® proves to be a comprehensive tool capable of significantly contributing to various stages of content production5-6. Literature also presents other possibilities for its application, such as in nursing education7 and qualitative data analysis8.
Despite these advantages, caution is needed when using ChatGPT®. This is because it is not suitable to create new texts based entirely from a prompt intended to be included in a manuscript. It should be noted that ChatGPT® is not an author, and therefore, attempts to use it as such may result in plagiarism of text generated by this tool. Some journals adopt resources to detect AI-generated texts, such as Originality® (https://originality.ai). Additionally, it is essential to note that ChatGPT® may make up references, sometimes presenting seemingly plausible but incorrect and/or superficial arguments. Consequently, the generated content may contain inaccuracies, making it crucial to thoroughly review the material to ensure accuracy5-6,9.
As technology advances, threats and challenges to its usage emerge, leading to the development of new tools to enhance and improve the utility of existing ones. One such tool developed to compensate for these ChatGPT® disadvantages is ScholarAI® (https://scholarai.io). This is a plugin developed to provide users with access to a database of peer-reviewed papers and academic research. ScholarAI® connects the large language models that power ChatGPT® with personalized access to open-access papers. Users can directly consult relevant peer-reviewed studies to find reliable information for scientific research, technical projects, and grant proposals.
Another AI program that has stood out is Bard®, created by Google AI in 2017. It uses an advanced language model capable of generating texts, translating languages, creating various types of content, and answering questions in an informative and accurate manner. In 2023, the Gemini® program (https://gemini.google.com) was developed, which is a family of language models from Google AI that includes Bard, representing an advancement in large language model technology. Notably, both programs are updated daily with current information.
Other tools have also proven useful in increasing researchers' productivity, such as Research Rabbit®, Paperpal®, Grammarly®, Scite®, R Discovery®, and SciSpace®. These platforms offer support at various stages of research, from organizing and analyzing references to improving writing and reviewing papers, providing a more efficient environment for academic production10.
Research Rabbit® (https://www.researchrabbit.ai) is a free tool to analyze networks of scientific papers, making the search for studies related to a specific topic easier and providing a broader understanding of the origins and evolution of debates. It also helps identify authors who collaborate in the same area. This AI recommends similar researches based on the user's collections, which keeps them updated. Additionally, its integration with Zotero, a free software to manage and share bibliographic references, allows users to extract references and citations in a practical way.
Both Paperpal® (https://paperpal.com) and Grammarly (https://www.grammarly.com) are tools for researchers and writers seeking to improve the quality of their texts. These platforms offer suggestions not only to correct grammatical errors but also to improve style, clarity, and content flow.
The Scite® platform (https://scite.ai/), in turn, goes beyond simple article searches as it offers context by showing how works have been cited by other authors. This provides a broader view of the impact of a specific article and proves especially useful when writing the discussion section of a study due to the detailed information on previously addressed concepts and ideas it provides.
With a database of approximately 100 million articles, R Discovery® (https://discovery.researcher.life) uses a recommendation system similar to social media by creating personalized feeds based on the user's areas of interest. Although it does not replace a complete literature review, it is a practical tool to stay up to date in daily research routines.
Another tool that can aid to understand academic texts better is SciSpace® (https://scispace.com). It acts as an intelligent reading assistant, allowing users to upload PDFs to identify key messages, explain complex concepts, and synthesize information. Some features include result summarization, description of tables and equations, and paraphrasing in different styles.
With such a number of resources available for scientific production, discussions about the limits of artificial intelligence have intensified, aiming to ensure that its use does not compromise scientific quality and integrity. In order to understand the position of major scientific publishers worldwide on the use of AI in the elaboration of papers, the guidelines of publishers such as Elsevier, Taylor & Francis, Wiley, and SAGE were consulted. These guidelines vary, but some common principles emerge.
Elsevier allows the use of AI to improve the readability and language of the work but makes it clear that such technology cannot replace intellectual tasks such as writing the introduction, discussing results, drawing conclusions, or providing recommendations. It also prohibits the use of AI to generate images, except for color enhancements, as well as listing AI as a co-author*.
Taylor & Francis allows the use of AI to improve language but advises that it must not be included in the authorship list and requires that its use be declared in detail. This includes specifying which tool was used, how it was used, in which section of the article, and why it was employed. It also requires that authors verify the authenticity of the information and check the article for plagiarism**.
Wiley takes a similar approach and allows the use of AI to improve writing but requires a detailed declaration of its use, recommending authors to verify the text for accuracy and plagiarism. It also highlights the need to declare AI use as a limitation of the study, pointing out potential problems arising from its application***.
Finally, SAGE adopts a similar position, allowing the use of AI to improve the text, provided it does not replace intellectual tasks. It also prohibits that AI be included in the authorship list and requires that its use be explicitly stated, including the inputs provided by the authors****.
In summary, although there is growing acceptance of the use of AI for language enhancement, all publishers emphasize the importance of maintaining human control over intellectual activities and transparently declaring the use of AI tools in the drafting of articles. This position aligns with the "Guide for the Use and Reporting of Artificial Intelligence in Scientific-Academic Journals"11 and the recommendations of the International Committee of Medical Journal Editors12, which provide guidelines for responsible AI use in scientific publishing. Among these guidelines, the recommendation that while AI can be used to improve aspects such as writing quality, formatting, and data analysis, its use should not replace essential intellectual work such as hypothesis formulation, result interpretation, and conclusion building. The main recommendations are:
-
Transparency: Authors must declare the use of AI in the manuscript in detail, specifying which tools were used, how, and at what stages of the process.
-
Intellectual Property: AI cannot be listed as a co-author since it does not perform intellectual or creative activities.
-
Verification and authenticity: Authors are responsible for ensuring the originality of the work and checking for potential plagiarism or inaccuracies generated by AI.
-
Limitations: AI use should be considered a limitation of the study, and potential problems arising from its application should be discussed in the publication.
Similarly, the Association for Computational Linguistics (ACL) recognizes the use of artificial intelligence as a valid tool for both linguistic assistance and literature search in the preparation of academic papers. Regarding language assistance, when generative models are used exclusively to enhance original content without generating new material, their use is considered similar to widely accepted tools such as spell checkers and synonym dictionaries, which have long been an integral part of the academic writing process. In relation to literature search, AI can serve as an efficient assistant for locating and identifying relevant references. However, researchers must preserve the same standards of rigor regarding the accuracy of citations and the depth of the literature review. Additionally, it is essential to be cautious of potential biases in citations suggested by generative algorithms13.
These guidelines support the idea that the use of AI is acceptable as long as it is employed as technical and linguistic support, preserving the intellectual integrity of the research and writing process. Given that AI is a constantly evolving technology, it will be essential to continuously review and update recommendations and best practices. The constant technological evolution demands that the editorial body frequently revisits its training and qualifications in editing to update internal policies with a special focus on ethics, morality, and scientific quality, ensuring the rigor and integrity of academic production11.
In this perspective, Revista Texto e Contexto Enfermagem acknowledges the potential of AI as a tool to support scientific production and publication. This implies that editorial actions must align with current national and international recommendations, ensuring ethics and morality in the editorial process in a transparent, responsible manner, with a focus on maintaining high quality.
Consequently, even though the topic may be further developed, it can be stated that AI tools applied to scientific writing can be useful for some tasks, optimizing researchers' time. It should be clarified that they do not replace the need for profound theoretical-methodological knowledge nor do they dispense with the intellectual work that is the essence of the creator, researcher, and producer, the author. Therefore, these tools should be viewed not as definitive solutions or threats but as instruments whose potential should be explored, understood, and critically used by researchers.
REFERENCES
-
1. Anyoha R. The History of Artificial Intelligence. Sci News [Internet]. 2017 [cited 2024 Sep 04]. Available from: https://sitn.hms.harvard.edu/flash/2017/history-artificial-intelligence/
» https://sitn.hms.harvard.edu/flash/2017/history-artificial-intelligence/ -
2. Bini SA. Artificial Intelligence, machine learning, deep learning, and cognitive computing: What do these terms mean and how will they impact health care? J Arthroplasty [Internet]. 2018 [cited 2024 Sep 04];33(8):2358-61. Available from: https://doi.org/10.1016/j.arth.2018.02.067
» https://doi.org/10.1016/j.arth.2018.02.067 -
3. Inglada Galiana L, Corral Gudino L, Miramontes González P. Ethics and artificial intelligence. Rev Clin Esp (Engl Ed) [Internet]. 2024 [cited 2024 Sep 04];224(3):178-86. Available from: https://doi.org/10.1016/j.rceng.2024.02.003
» https://doi.org/10.1016/j.rceng.2024.02.003 -
4. Fabrizzio GC, Oliveira LM, Costa DG, Erdmann AL, Santos SLG. Assistente virtual: ferramenta para produção em saúde no enfrentamento à COVID-19. Texto Contexto Enferm [Internet]. 2023 [cited 2024 Sep 04];e32. Available from: https://doi.org/10.1590/1980-265X-TCE-2022-0136p
» https://doi.org/10.1590/1980-265X-TCE-2022-0136p -
5. Fuchs K. Exploring the opportunities and challenges of NLP models in higher education: Is Chat GPT a blessing or a curse? Front Educ [Internet]. 2023 [cited 2024 Sep 06];8. Available from: https://doi.org/10.3389/feduc.2023.1166682
» https://doi.org/10.3389/feduc.2023.1166682 -
6. Hill-Yardin EL, Hutchinson MR, Laycock R, Spencer SJ. A Chat (GPT) about the future of scientific publishing. Brain Behav Immun [Internet]. 2023 [cited 2024 Sep 04];110:152-4. Available from: https://doi.org/10.1016/j.bbi.2023.02.022
» https://doi.org/10.1016/j.bbi.2023.02.022 -
7. Sasso GTMD, Lanzoni GMM, Alvarez AG, Barra DCC, Barbosa SFF. Potential contribution of CHATGPT® To Learning About Septic Shock In Intensive Care. Texto Contexto Enferm [Internet]. 2024 [cited 2024 Sep 04];33:e20230184. Available from: https://doi.org/10.1590/1980-265X-TCE-2023-0184en
» https://doi.org/10.1590/1980-265X-TCE-2023-0184en -
8. Zhang H, Wu C, Xie J, Lyu Y, Cai J, Carroll JM. Redefining qualitative analysis in the AI era: Utilizing ChatGPT for efficient thematic analysis - version 3. arXiv preprint [Internet]. 2023 [cited 2024 Sep 06];arXiv:2309.10771. Available from: https://arxiv.org/abs/2309.10771
» https://arxiv.org/abs/2309.10771 -
9. Grzybowski A, Pawlikowska-Łagód K, Lambert WC. A History of Artificial Intelligence. Clin Dermatol [Internet]. 2024 [cited 2024 Sep 04];42(3):221-9. Available from: https://doi.org/10.1016/j.clindermatol.2023.12.016
» https://doi.org/10.1016/j.clindermatol.2023.12.016 -
10. Jhajj KS, Jindal P, Kaur K. Use of Artificial Intelligence tools for research by medical students: a narrative review. Cureus [Internet]. 2024 [cited 2024 Sep 04];16(3):e55367. Available from: https://doi.org/10.7759/cureus.55367
» https://doi.org/10.7759/cureus.55367 -
11. Penabad-Camacho L, Morera-Castro M, Penabad-Camacho MA. Guide for the use and reporting of Artificial Intelligence in scientific-academic journals. Rev Eletr Educare [Internet]. 2024 [cited 2024 Sep 6];28(S):1-41. Available from: https://www.revistas.una.ac.cr/index.php/EDUCARE/article/view/19830
» https://www.revistas.una.ac.cr/index.php/EDUCARE/article/view/19830 -
12. International Committee of Medical Journal Editors (ICMJE). Recommendations for the Conduct, Reporting, Editing, and Publication of Scholarly Work in Medical Journals [Internet]. 2024 [cited 2024 Sep 04]. Available from: https://www.icmje.org/icmje-recommendations.pdf
» https://www.icmje.org/icmje-recommendations.pdf -
13. Spinak E. Inteligência Artificial e a comunicação da pesquisa. SciELO em Perspectiva [Internet]. 2023 [cited 2024 Sep 10]. Available from: https://blog.scielo.org/blog/2023/08/30/inteligencia-artificial-e-a-comunicacao-da-pesquisa/
» https://blog.scielo.org/blog/2023/08/30/inteligencia-artificial-e-a-comunicacao-da-pesquisa/
