Artificial intelligence in nurse education – a new sparring partner?

In Norway, one of four nursing students are failing on the national exam for nursing students in the subject of anatomy, physiology, and biochemistry.We need more knowledge on how to address this situation and this position paper builds on our previous work on AI in education and examines GPT-4's performance on a full-scale exam within nursing (multiple choice exam).The primary data sources were GPT-4's ability to answer correctly/incorrectly on 53 exam questions within the National exam in Anatomy, Physiology, and Biochemistry in nursing education in Norway.From the same exam, a further multimodal analysis was carried out to examine GPT-4's ability to answer correctly/ incorrectly on 5 of the exam questions based on visual illustrations (blinded answers).The findings from this intrinsic case study revealed that GPT-4 performs very well on this summative exam with an accuracy rate ranging from 84.9% to 94.5% (based on three test scenarios), which indicates a strong understanding of the exam questions in this national nursing exam.This, and its ability to answer correctly on the five multimodal analysis-questions (62,5%-100% correct) shows both the GPT-4's ability to handle Norwegian exam language and its multimodal capabilities.The (digital) fieldwork after the exam-testing shows that GPT-4 can enhance the possibilities of formative assessment by providing timely and personalized feedback, which supports student learning.For summative assessment, GPT-4 demonstrated reliable evaluation of student exams on multiple-choice questions, but also on some process questions as part of the exam.One implication of the study is that GPT-4 can be a sparring partner for nursing students in formative assessment processes, case simulations, and so on, since it shows high reliability on national exams within Anatomy, Physiology and Biochemistry.GPT-4 can scaffold tutoring, ZPD, and formative and summative assessment practices.However, the potential for perpetuating biases and ethical issues underscores the need for vigilance and careful implementation of such AI.