Using chatbots like <scp>ChatGPT</scp> to support nursing practice

Abstract

For the past several months, there has been increased interest in the possibilities of using artificially intelligent (AI) chatbots to help craft educational, clinical and scientific documents following the November 2022 realise of the open-source platform, ChatGPT. ChatGPT is a large language model developed by OpenAI that uses machine learning techniques to generate human-like text. It is based on the GPT (Generative Pre-trained Transformer) architecture, which uses deep learning to analyse and understand natural language text. ChatGPT is certainly not the first chatbot supporting healthcare professionals. Xu et al. (2021), in their review of chatbots used in cancer care identified 78 chatbots that were applied to support diagnosis, treatment, monitoring, workflow planning and health promotion. However, unlike previous chatbots that had limited capabilities and specific use, ChatGPT has been trained on a diverse range of texts including web pages, books and conversational data. Consequently, it is highly versatile and can be used for a wide range of areas, including customer support, information retrieval, education and health care (OpenAI, 2022). We decided to examine the platform by asking the AI platform itself (https://chat.openai.com/auth/login) the following questions related to its potential use in nursing practice: Describe the potential use of ChatGPT in nursing practice. Identify the benefits and the limitations. What are the implications of using ChatGPT? The response is shown in Table 1. The generated automated response highlights several possible applications that could support nurses in their clinical practice. According to ChatGPT, it could reduce repetitive writing and administrative work such as summarising long lists of patient information. It could provide case summaries or care plans identifying nursing interventions targeting specific patient needs. It could enhance communication by providing conversation cues between nurses and patients, and it could generate instructions and recommendations that are easier to follow, jargon free and person/patient centred. Other potential applications could be translating medical language into easier to understand text or translating information/instructions directly to the patients' native language. Moreover, complex instructions could be easily and efficiently simplified, thereby possibly increasing patient compliance and adherence. Obviously, these benefits make its use appealing. However, its use is not without limitations and risks. Compassionate and empathic communication is the bases of the nurse–patient relationship. It is possible that overreliance on these chatbots can lead to deskilling nurses. For example, providing prescriptive responses to nurse–patient conversations may make these interactions more impersonal and less therapeutic. Such a concern is supported by Parviainen and Rantala (2022) who argued the use of AI chatbots to provide automated consultations and decision-making can have a profound influence on the nurse–patient relationship, especially regarding its effect on patients' trust. Thus, important questions to consider in relation to its use include ‘Will the patient's trust towards nurses be eroded if they perceive that decisions are being taken by chatbots rather than by human beings?’ ‘Will patients adhere to the recommendations suggested by nurses, if they notice that the decisions are being supported by chatbots?’. Responses given by these chatbots may not be reliable or evidence based. In fact, the OpenAI website acknowledges ChatGPT ‘may occasionally generate incorrect or misleading information and produce offensive or biased content. It is not intended to give advice’. However, as with other chatbots, there is the risk that technologically smart nurses may use this tool without considering its limitations. Doing so could increase the risk of giving inaccurate or biased information to patients or other staff. Since AI robots cannot take