ChatGPT took the world by storm when it was released to the public in November 2022. Only 2 months after it was released, it has more than 30 million users and got roughly 5 million visits a day. This made it one of the fastest growing software products in history. To put that into perspective, it took Instagram nearly a year to get 10 million users.1
ChatGPT, which stands for Chat Generative Pretrained Transformer, is an advanced, pretrained, deep learning language model developed by OpenAI, an artificial intelligence (AI) research laboratory based in San Francisco, California. It is a deep learning model that uses a neural network (a type of AI model designed to mimic the way the human brain processes information through interconnected artificial neurons) to generate natural language text in response to a given query.2 Using a technique called the “Transformer architecture,” ChatGPT is able to process vast amounts of information and generate large amounts of coherent and informative text in real time, based on simple, even short, vague, or ambiguous prompts. This makes it a popular choice for conversational AI applications, such as language models and chatbots, whose goal is to generate text similar to human speech.3,4
However, ChatGPT's expeditious growth has also sparked concerns in the academic community, particularly when news broke out that the New York Department of Education suddenly announced a ban on access to the program due to concerns of academic integrity violations.5,6 This development, along with an exponential rise in reports of students caught cheating on papers and examinations, has led to a bombardment of discussions about the impact of ChatGPT on education.7–9 Furthermore, ChatGPT has demonstrated the ability to pass various professional and licensing examinations, including the United States Medical Licensing Examination (USMLE),10 the Wharton MBA final examination,11 and components of the bar examination,12 further intensifying the ethical concerns surrounding its use. In the face of dire warnings of plagiarism, breaches of academic integrity, theft of intellectual property, and troves of ethical concerns, it is of utmost importance for faculty to be cognizant of the potential advantages, benefits, limitations, and drawbacks associated with the use of technology in the academic setting.
Despite the ethical concerns surrounding its use, ChatGPT has demonstrated a remarkable range of extremely useful capabilities in the realm of language generation and processing of vast amounts of information. This has already led to its integration into various industries. It is extremely important for nursing faculty to understand the potential applications, limitations, and drawbacks of ChatGPT in order to make informed and effective decisions about its integration into nursing curricula and coursework. This thorough examination will allow faculty to create curricula and coursework that not only promotes academic integrity but also fosters independent thinking and critical reasoning among students.
What Can It Currently Do?
Rapid assistance: ChatGPT can provide students with immediate responses to a multitude of queries, covering a vast range of topics, content, and information. This can be an invaluable tool in enhancing students' grasp of complex concepts, such as pathophysiology, pharmacology, diagnostic reasoning, and critical thinking. With additional prompts, it can efficiently outline rationales for decision making in complex clinical scenarios. These features can all be immensely helpful to the student.13,14 Supplemental Digital Content Figure 1, available at: https://links.lww.com/NE/B294, provides an example of a critical analysis of the Quintuple Aim and strategies for future nurse practitioners (NPs) to use in leading the efforts to meet the aim.
Rapid resource accessibility: ChatGPT's ability to process and analyze vast amounts of information provides students a unique opportunity to receive immediate answers to questions and be promptly directed toward relevant resources. This capability can be a valuable tool for students to quickly access and utilize information, allowing increased efficiency in their studies. The immediate access can greatly enhance students' learning experiences and save valuable time and effort. This rapid access has the potential to greatly benefit students in their academic pursuits (see Supplemental Digital Content Figure 1, available at: https://links.lww.com/NE/B294).
Writing assistance: ChatGPT's most remarkable attribute is its exceptional efficiency and effectiveness in generating seemingly high-quality written responses. It can rapidly paraphrase text, generate ideas and outlines, expand and expound on related concepts, provide a structure framework for organizing ideas, and generate well-crafted introductions and conclusions, among a myriad of other writing capabilities. Unlike traditional tools such as Rephrase, Wordtune, Scribbr, Grammarly, QuillBot, and so forth, ChatGPT represents a much more advanced form of technology. However, it should be noted that although the generated text might sound good to an unsuspecting reader, the information may be inaccurate, outdated, or even fabricated.2 It should also be noted that while ChatGPT may be able to generate references or intext citations, the authors found that most of its cited sources were not valid, nor were they retrievable when cross-referenced. (see Accuracy and Citations and References later in the article for additional discussion) (see Supplemental Digital Content Figure 1, available at: https://links.lww.com/NE/B294).
Research support: ChatGPT has the ability to assist students in research endeavors by providing relevant information and data on various topics of interest. It can efficiently produce key resources related to the topic, offer comprehensive summaries of previous studies, and provide recommendations for further research. This type of support enables students to effortlessly expand their knowledge and understanding of the topic at hand. However, as mentioned previously, the veracity and reliability of its output must be verified.2 It is also important to note that there has been some opposition from prominent journals regarding the use of ChatGPT as an author15 (see Supplemental Digital Content Figure 1, available at: https://links.lww.com/NE/B294).
Examination preparation/study material generation: ChatGPT has the capability to generate outlines, study guides, and study questions to aid students in their studying. It can create open-ended questions for critical thinking and diagnostic reasoning, as well as provide detailed responses with rationale for each question. In addition, it can also produce multiple-choice questions with the correct answer and accompanying rationale. These features can be an extremely valuable tool for students to help reinforce their understanding of course material and to better prepare them for examinations. These features can also serve as an incredibly useful resource for faculty in creating study guides and examination questions. Supplemental Digital Content Figure 2, available at: https://links.lww.com/NE/B295, provides 3 multiple-choice questions similar to what could be found on an NP certification examination.
Examination question response/rationale: This is perhaps one of the most concerning features of ChatGPT. It has already demonstrated its ability to pass the USMLE examination, which features critical thinking and diagnostic reasoning questions, similar to questions in nursing licensure and certification examinations. This poses a concern for distance education programs whose students may have easy access to ChatGPT in a manner that may not be easily detected by current remote examination proctoring systems (see Supplemental Digital Content Figure 2, available at: https://links.lww.com/NE/B295).
Proofreading and editing support: ChatGPT is a formidable proofreading and editing tool. It can effectively and efficiently identify and rectify grammatical errors, awkward sentence structures, stilted syntax, and incorrect spelling or punctuation. In addition, it can assist with word choice, suggest alternative phrasing, and help maintain a consistent tone and style in the student's writing, ensuring a preferred voice and presentation. Overall, ChatGPT serves as an extremely valuable and expedient resource for students seeking to improve the quality of their writing.
Summarization assistance: ChatGPT is a powerful tool that uses sophisticated algorithms for key word extraction, sentence scoring, text compression, and advanced language comprehension capabilities to generate succinct yet comprehensive summaries of lengthy articles, texts, or documents. This can be a useful tool for students who need to quickly navigate through extensive materials. Moreover, ChatGPT has advanced sentiment analysis capabilities, which can further help contribute to its effectiveness in summarizing text and understanding the emotions and opinions expressed in the content. Although this can be a valuable study tool, it also raises concerns regarding potential academic misconduct, such as students' utilization of these generated summaries and passing them off as their own, while bypassing the actual work needed to produce it.
Clinical documentation generation: ChatGPT has the capability to produce subjective, objective, assessment and plan (SOAP) notes and complete full history and physical examinations (H&Ps). As with any of its generated responses, it will initially provide a concise and succinct output that can be further developed and expanded upon through further questioning and more detailed querying. Supplemental Digital Content Figure 3, available at: https://links.lww.com/NE/B296, provides an example of a SOAP note and a detailed history and physical (H&P).
What Are Its Limitations?
Accuracy: As mentioned earlier, a significant limitation of ChatGPT is its accuracy. It is imperative for students and faculty to understand that, while it may offer efficiency and time-saving benefits, it is extremely susceptible to inaccuracies and inconsistencies. To effectively utilize ChatGPT, it is necessary to possess sufficient knowledge to detect any errors and to exercise due diligence in verifying the accuracy of its output.3,16
Citations and references: ChatGPT is limited in its ability to generate credible references and citations. In some instances, it will quickly list out citations of references, while in other instances, it responds that it is an AI language model without access to the internet and/or specific references to the information provided. In many instances, even when it produces references in response to specific queries, the citations have been inaccurate as it is not equipped with the ability to verify the accuracy or the authenticity of the information it presents. This can lead to the presentation of incorrect or misleading information, which is especially concerning in academic settings where credibility and accuracy are of utmost importance. There have been instances in which the program appeared to provide legitimate references in response to the author's query, only to retract the information and admit to being mistaken upon further inquiry. However, ChatGPT directs the user to specific databases, such as PubMed and Cumulative Index to Nursing and Allied Health Literature (CINAHL), and professional organizations such as the National Organization of Nurse Practitioner Faculties (NONPF) and the American Association of Colleges of Nursing (AACN), as well as general resources such as conferences, workshops, and online communities, where one can find additional information on resources on the topic of inquiry. Readers should see Supplemental Digital Content Figure 1, available at: https://links.lww.com/NE/B294, which provides an example of a critical analysis of the Quintuple Aim and strategies for future NPs. Note, however, that the references generated are all inaccurate.
Limited up-to-date information: The current iteration of ChatGPT has a knowledge base cutoff date of 2021, which limits its capability to provide, summarize, or synthesize accurate and current information. Nursing and health care are rapidly evolving, and any recent developments or advancements may not be incorporated or reflected in ChatGPT's responses, which may perpetuate outdated or inaccurate information. It is important to note that though language models are continually being updated and improved, they are still limited by the training data.2
Contextual understanding: ChatGPT's strength lies in its ability to effortlessly generate responses through analysis of vast amounts of information and advanced pattern recognition. Unfortunately, it may not always have sufficient context to fully comprehend the nuances of a question or to provide a fully accurate answer. This limitation becomes apparent when attempting to use ChatGPT for complex case study write-ups or analyses and discussion of patients seen in the clinical setting.
Medical terminology/jargon: Despite ChatGPT's vast database that is well versed in medical terminology, it does have limitations in its ability to fully comprehend medical jargon or technical terms. It is possible for ChatGPT to misinterpret medical terms or use them in the wrong context. However, as ChatGPT was designed for continuous training with ongoing incorporation of user feedback, it is unclear whether any inaccuracies currently exist. The authors were not able to detect any inaccuracies in our use. ChatGPT is capable of interpreting shorthand and acronyms commonly found in medical records and can provide detailed information on recent medical phenomena, such as long-haul COVID-19.
Exam question generation: Although ChatGPT is able to generate questions for study and examination preparation, the authors found that there were limitations with regard to the types of questions it can produce. Although it can generate low-level knowledge-type multiple-choice questions and provide detailed answers with rationale to more complex questions, its ability to actually generate higher-level critical thinking–type multiple-choice questions appears much more limited. This is likely because critical thinking and diagnostic reasoning questions often require a deeper understanding of the subject matter and a more nuanced approach to problem solving. These types of questions require more sophisticated algorithms and a deeper understanding of the context in which they are being asked. As a result, ChatGPT may struggle to generate these types of questions, though it can correctly answer them and provide rationale. It is important for students and faculty to understand these limitations and use ChatGPT question generation capabilities accordingly (see Supplemental Digital Content Figure 2, available at: https://links.lww.com/NE/B295, which provided 3 multiple-choice questions similar to what could be found on an NP certification examination).
Bias: The limitations of ChatGPT stem from the data that it is exposed to. If the original data contain biases, it is highly likely that ChatGPT's responses and output will continue to perpetuate those biases. This raises significant ethical concerns with the use of ChatGPT and AI technology, as bias in training data can significantly impact the performance, output, and accuracy of these types of programs. There is a growing body of evidence and literature on the issue of bias in AI. Many studies have shown that AI algorithms can perpetuate and amplify existing biases.17 Many studies have shown that sources of bias in AI algorithms can further perpetuate health care disparities.18
Lack of personal reflection: Because ChatGPT is a language model trained to generate text based on algorithms, patterns, and relationships in data, it does not have personal experiences or emotions. It cannot provide reflections or insights based upon personal experiences or emotions. It is nearly impossible for ChatGPT to create a personal reflection that is specific to the student or the patient. Supplemental Digital Content Figure 4, available at: https://links.lww.com/NE/B297, provides an example of a family assessment and a cultural assessment correlating to the H&P that was previously generated and shown in Supplemental Digital Content Figure 3, available at: https://links.lww.com/NE/B296.
Nontext information analysis: ChatGPT, being a language model designed to process and generate text-based responses, cannot process any type of nontext input, such as images, videos, or audio. This severely limits its capabilities to understand and provide responses to certain types of questions that contain a visual or audio component. ChatGPT cannot discuss, analyze, or summarize any nontext input.
Word limit: ChatGPT has a word limit for the input it can process and it is typically equivalent to 3000 to 4000 words. The limitation is actually something called a “token limit,” which is based upon a combination of the number of characters and words and its formatting structure. The token limit can also vary depending on the specific implementation of ChatGPT. This limits the user from being able to input entire articles or book chapters and requires the user to break down a large input into smaller sections.
Acceptance as a source: It is important to note that there has been opposition from prominent journals regarding accepting the works written by or generated by AI language models such as ChatGPT. These concerns stem from beliefs that authorship should be limited to human beings who generate the innate creative and intellectual contributions that are not found in AI programs. In addition, the challenges with verifying all of ChatGPT's potential inaccuracies and biases may tarnish and compromise the integrity of the publication. This highlights the need for caution and careful consideration in the utilization of ChatGPT in academic research.15
Therefore, it is important for both students and faculty to be aware of the limitations and biases that may exist within ChatGPT and to approach use with a critical eye. Ultimately, while ChatGPT can be an immensely advanced, powerful, and time-saving tool, it is not foolproof and may contain inconsistencies, inaccuracies, and undisclosed biases. Although it can be used to support many of the activities found in academia, it still requires a knowledgeable and diligent human overseer to ensure its accuracy and appropriate use.
Recommendations for Faculty
- Educate yourself and your students on the limitations and potential inaccuracies, biases, and ethical concerns with the use of ChatGPT. Ultimately, while this can be an incredibly useful tool, it is important for students and faculty to recognize its limitations and approach its use with caution.3
- Establish clear guidelines for use. Be firm and clear with consequences of any plagiarism or other violations of academic integrity.19,20
- Establish consensus and determine appropriate usage of ChatGPT applications in assignments.21,22
- Encourage faculty to explore the capabilities and applications of ChatGPT.23,24 Utilize it as a tool to validate students' work and cross-reference their writing. There are free tools available online to help detect AI-written work, such as AI Writing Check and OpenAI's AI Text Classifier. AI Writing Check requires a 100-word minimum sample, while OpenAI Text Classifier requires a minimum of 750 characters for analysis. Both tools warn that they are not entirely accurate, and students can often outsmart the system by rephrasing the AI-generated text. Ironically, ChatGPT may not consistently nor accurately identify its own output, especially if the text has been rephrased or altered in any way.
- Create assignments that require intricate thinking, comprehensive context, and self-directed learning to decrease students' dependence on ChatGPT. An example would be to create a given case scenario and require students to provide specific and individualized responses in their discussion and analyses. Limiting the entire class to a single case scenario discourages students from relying solely on duplicated responses generated by ChatGPT, thus fostering independent thinking and creativity.
- Design assignments that promote verbal expression, such as synchronous presentations, live discussions, or debates (either in-person or online). This will reduce the reliance on ChatGPT for completing the required deliverables.
- Develop examinations that minimize the use of a multiple-choice format and incorporate examination questions that require image interpretation or open-ended responses to patient scenarios. This approach is especially important for distance education programs that do not have on-site proctored examinations or limitations in examination-proctoring software capabilities.
- Construct assignments that focus on visual or multimedia elements, such as videos, graphics, or images, and require students to engage in extensive reflection, discussion, analysis, and interpretation of the content. This will minimize students' overreliance on ChatGPT for assignments, as it is unable to process nontextual input.
- Use simulation scenarios to evaluate and test students' competencies in patient assessment, diagnosis, and treatment in real-world, real-time clinical settings. This approach encourages students to apply their knowledge and skills and to engage in critical thinking “on the spot.”
- Design collaborative group projects that reduce dependence on individually written deliverables. Create projects that require students to incorporate collaboration, communication, teamwork, and problem-solving skills. This type of assignment prompts students to work together, share their perspectives, and learn from one another, instead of relying solely on ChatGPT-generated responses. The uniformity of ChatGPT's answers can make it challenging for a group to generate unique, personalized, and diverse ideas and responses.
ChatGPT is a highly advanced language model that has the potential to revolutionize nursing education by elevating the way students interact with technology, offering students quick and convenient access to information, and enhancing students' abilities to generate text-based information. Its ease of use, speed, and ability to generate text-based information makes it an extremely valuable tool for many applications. However, there are valid concerns regarding academic integrity, as students may be tempted to use the model to cheat on assignments or examinations. In addition, there is the risk of students becoming overly reliant on this technology, which could lead to a decline in original thought, independent learning, and critical thinking skills.
To mitigate these risks, it is imperative that faculty establish clear guidelines and policies for ethical use. Students must be taught how to critically evaluate information obtained from ChatGPT and other sources, in order to make informed decisions in their academic and professional lives. To minimize students' reliance on ChatGPT, faculty should create assignments and examinations that emphasize complex problem solving, profound reflection, synergistic collaboration, independent thought, and self-directed learning. Moreover, it is crucial that faculty themselves become familiar with the program and its capabilities and limitations to minimize its improper use, effectively incorporate it into their teaching strategies, and more accurately assess student competency. Integrating ChatGPT (or any other chatbot/AI application) into nursing education presents both opportunities and challenges. It is up to the educators and students to navigate these challenges and make the most of this innovative tool. As the Nobel Peace Prize winner Christian Lous Lange once said, “Technology is a useful servant but a dangerous master.”25 By using ChatGPT in an informed and responsible manner, nursing education can reap its benefits while avoiding its potential pitfalls. Furthermore, incorporating a critical analysis of the use of technology in nursing education can promote ethical and responsible use of technology, while also preparing students for the rapidly evolving technological landscape in health care.
1. Roose K. How ChatGPT kicked off an AI arms race. New York Times. Published February 3, 2023. Accessed February 3, 2023. https://www.nytimes.com/2023/02/03/technology/chatgpt-openai-artificial-intelligence.html
2. OpenAI. ChatGPT: optimizing language models for dialogue. Published 2022. Accessed January 30, 2023. https://openai.com/blog/chatgpt/
3. Lund BD, Wang T. Chatting about ChatGPT: how may AI and GPT impact academia and libraries. Library Hi Tech News. 2023. doi:10.2139/ssrn.4333415
4. Dale R. GPT-3: what's it good for? Nat Lang Eng. 2021;27(1):113–118. doi:10.1017/S1351324920000601
5. Rosenblatt K. ChatGPT banned from New York City public schools' devices and networks. NBC News. Published January 5, 2023. Accessed January 30, 2023. https://www.nbcnews.com/tech/tech-news/new-york-citypublic-schools-ban-chatgpt-devices-networks-rcna64446
6. Korn J, Kelly S. New York City public schools ban access to AI tool that could help students cheat. CNN. Published January 6, 2023. Accessed January 30, 2023. https://www.cnn.com/2023/01/05/tech/chatgpt-nycschool-ban/index.html
7. BBC News. What is ChatGPT, the AI software taking the internet by storm? [Video]. YouTube. Published January 15, 2023. Accessed January 30, 2023. https://www.youtube.com/watch?v=BWCCPy7Rg-s
8. Mitchell A. Professor catches student cheating with ChatGPT: ‘I feel abject terror’. New York Post. Published December 26, 2022. Accessed January 30, 2023. https://nypost.com/2022/12/26/students-using-chatgpt-tocheat-professor-warns/
9. Levin J. Student caught cheating using AI on final; Academic integrity policy updated. The Commentator. Published January 5, 2023. Updated January 6, 2023. Accessed January 6, 2023. https://yucommentator.org/2023/01/students-caught-cheating-using-ai-on-final-academic-integrity-policy-updated/
10. Kung TH, Cheatham M, Medenilla A, et al. Performance of ChatGPT on USMLE: potential for AI-assisted medical education using large language models. medRxiv. Published December 21, 2022. Accessed January 30, 2023. doi:10.1101/2022.12.19.22283643
11. Terwiesch C. Would ChatGPT Get a Wharton MBA? A prediction based on its performance in the Operations Management Course. Mack Institute for Innovation Management at the Warton School, University of Pennsylvania. Published 2023. Accessed January 30, 2023. https://mackinstitute.wharton.upenn.edu/wp-content/uploads/2023/01/Christian-Terwiesch-Chat-GTP-1.24.pdf
12. Bommarito M, Katz D. GPT takes the bar exam. SSRN. Published December 29, 2022. Accessed January 30, 2023. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4314839
14. Baker RS. Artificial intelligence in education: bringing it all together. In: OECD Digital Education Outlook 2021: Pushing the Frontiers With Artificial Intelligence, Blockchain and Robots. OECD Publishing; 2021:43–54.
15. Reed B. Science journals ban listing of ChatGPT as co-author on papers. The Guardian. Published January 27, 2023. Accessed January 30, 2023. https://www.theguardian.com/science/2023/jan/26/science-journals-ban-listing-of-chatgpt-as-co-author-on-papers
16. Biswas S. ChatGPT and the future of medical writing. Radiology. Published February 2, 2023. Accessed February 10, 2023. https://pubs.rsna.org/doi/full/10.1148/radiol.223312
17. Leavy S, O'Sullivan B, Siapera E. Data, power and bias in artificial intelligence. arXiv:2008.07341. Published July 28, 2020. Accessed February 10, 2023. https://arxiv.org/pdf/2008.07341.pdf
18. Celi L, Cellini J, Charpignon M, et al. Sources of bias in artificial intelligence that perpetuate healthcare disparities. PLOS Digital Health. 2022;2(3):e0000022. doi:10.1371/journal.pdig.0000022
19. Cotton D, Cotton P, Shipway R. Chatting and cheating: ensuring academic integrity in the era of ChatGPT [preprint]. Published January 10, 2023. Accessed February 10, 2023. https://edarxiv.org/mrz8h/
20. Hern A. AI bot ChatGPT stuns academics with essay-writing skills and usability. The Guardian. Published December 4, 2022. Accessed January 11, 2023. https://www.theguardian.com/technology/2022/dec/04/ai-bot-chatgpt-stuns-academics-with-essaywriting-skills-and-usability
21. Schwendicke F, Samek W, Krois J. Artificial intelligence in dentistry: chances and challenges. J Dent Res. 2020;99(7):769–774. doi:10.1177/0022034520915714
22. Gray K, Slavotinek J, Dimaguila GL, Choo D. Artificial intelligence education for the health workforce: expert survey of approaches and needs. JMIR Med Educ. 2022;8(2):e35223. doi:10.2196/35223
23. Robert N. How artificial intelligence is changing nursing
. Nurs Manage. 2019;50(9):30–39. doi:10.1097/01.NUMA.0000578988.56622.21
24. Lee D, Yoon S. Application of artificial intelligence-based technologies in the healthcare industry: opportunities and challenges. Int J Environ Res Public Health. 2021;18(1):271. doi:10.3390/ijerph18010271
25. Christian Lange. Nobel lecture. Published December 13, 1921. Accessed February 10, 2023. https://www.nobelprize.org/prizes/peace/1921/lange/lecture/