“Many nurses leave the profession because of all the heavy work involved and the wear and tear on the body,” said Pruinelli, who has worked extensively with machine learning and the adoption of intelligent systems. 

Recently, she co-authored an article on ChatGPT, the artificial intelligence chatbot, that concluded the technology holds promise for nursing education, but still needs work. 

The article, “The ChatGPT Effect and Transforming Nursing Education with Generative AI: Discussion Paper,” was published in Nurse Education in Practice and aimed to demonstrate the technology’s potential in assisting nurses with clinical documentation. 

UF has embraced AI, including launching an Artificial Intelligence Initiative geared toward setting the standard for an AI university. Administration hopes to prepare AI-capable students in every discipline taught at the university. As part of the university’s AI initiative, the College of Nursing has several recognized researchers focusing on AI, including Pruinelli. 

For Pruinelli’s article, researchers presented a prepared patient case study to ChatGPT, along with questions seeking nursing diagnoses, nursing interventions and nursing outcomes. 

“We tried to see if ChatGPT could do that kind of compilation of information and deliver the diagnoses,” Pruinelli said. 

While the program returned appropriate nursing diagnoses, most were not in line with the requested North American Nursing Diagnosis Association – International, or NANDA-I, classification, which uses standardized nursing terminology and diagnoses.

Along with the appropriate responses, however, results included inappropriate ones and a diagnosis not in the NANDA-I. Also, diagnosis codes were not correct.

“ChatGPT got a little bit lost and gave misinformation. That doesn’t mean training ChatGPT can’t give us better responses,” Pruinelli said. “The major problem is ChatGPT uses publicly available information, including not scientifically ‘proven’ evidence.”

While the chatbot can access copious amounts of information, it does not always return the most accurate information.

ChatGPT, short for Chat Generative Pre-training Transformer, is based on generative artificial intelligence and formulates responses to questions in human-like conversation. Users can ask follow-up questions. The chatbot draws from billions of parameters and millions of websites to create its responses. Through continuous human interaction, including feedback from users, the chatbot learns and improves.

While Pruinelli believes generative artificial intelligence will eventually assist nurses, it will not supplant them.

“The idea is not to replace what we do, but to use the technology to assist us or use it as an adjunct for our decision-making,” she said. “It doesn’t matter how accurate ChatGPT or any other technology is; we need to make sure it is actually the right intervention, the right diagnosis, the right recommendation for that patient, for that problem, at that time. The endpoint is always us.”