Artificial intelligence (AI) is an umbrella term for different data science approaches, including fuzzy expert systems, Bayesian networks, artificial neural networks, hybrid intelligent systems, and machine learning (Helm et al., 2020). Although AI has been in the research literature since the 1970s, the widespread use of electronic health records (EHRs) has created massive data sets ripe for extraction and analysis with advanced computational systems (Amisha et al., 2019). For example, Milinovich and Kattan (2018) estimated 32,000 discrete data elements in EHRs for every patient admission. Each person's admission contains inputs (presenting signs and symptoms, laboratory tests, diagnostic tests, working diagnoses, etc.) and known outcomes (the discharge diagnosis and treatment plan). Machine learning uses large sets of data to look for computational relationships (algorithms) between the inputs and outcomes. The computer system then "learns" over time when it encounters more data to improve its predictions of outcomes (Helm et al., 2020).

When researchers have confidence in predictions, health care providers can use the systems with new data to make predicted outcomes. AI has been used successfully in radiology (Chamberlin et al., 2022; Tanenbaum, 2020), oncology (Rezayi et al., 2022), obstetrics (Han & Datkheva, 2019), critical care (Hassan et al., 2021), and many other specialties. Though AI has demonstrated accuracy in making diagnoses, the judgment of experienced clinicians is necessary. For example, AI exploits data in EHRs to provide early warning for sepsis; however, patients with severe trauma with no sepsis can be flagged incorrectly because of the activation of neutrophils after tissue damage (Joosse et al., 2022). In this example, an experienced clinician would not start a sepsis protocol because it would not be indicated.

Swarm software offers a way to combine human intelligence in networked groups with AI technology to optimize well-reasoned professional judgments (Unanimous AI, 2022). The software platform, artificial swarm intelligence (ASI), is based on the swarm intelligence found in nature, for example, with flocks of birds or schools of fish that work in synchronicity to make decisions (Rosenberg & Willcox, 2020). An experiment at Stanford University tested the accuracy of ASI with eight radiologists who reviewed 50 chest x-rays using the Unanimous AI platform to determine the probability of pneumonia for each of the x-rays. The small team made 33 percent more correct diagnoses for pneumonia than AI alone or radiologists who reviewed the x-rays independently (Rosenberg & Willcox, 2020).

How could ASI be used in nursing education? Nursing faculty can collaborate with computer scientists to use existing AI systems and Unanimous AI systems to create patient cases that typically are difficult for nurse practitioner students to make a correct final diagnosis from differential diagnoses. In an ASI session, faculty can see where the nurse practitioner students lean toward a specific diagnosis and the level of conviction toward that diagnosis. This process can give faculty a unique insight into the thinking of students, reinforce correct diagnostic thinking, and correct misinterpretations in debriefing sessions. For more information, watch videos about ASI at https://youtu.be/xWSkbsIRNMg and https://www.youtube.com/watch?v=xODlyNdxuEY

REFERENCES