Thousands of New York City nurses spent weeks on the picket line this winter, demanding increased nurse-to-patient ratios and better safety protections in the largest nurses strike in the city’s history.

But one overlooked part of their demands is guidelines around artificial intelligence in medical settings. That’s one of the biggest technological developments and debates of the 21st century, once again raising the question, “what place does artificial intelligence have in a hospital?”

The answer is wide-ranging and changing every day. Doctors have already begun to use voice-transcription technology to cut down note-taking time, carry out administrative tasks, and use predictive models to look at possible patient health trajectories. A University of Minnesota study published in January 2025 found that 65% of hospitals already use AI-based tools. The number is likely higher today. 

Some hospitals are already using it as a diagnostic aid. But one well-publicized study showed that gastroenterologists in Poland got 20% worse at spotting polyps and other abnormalities on their own during colonoscopies, after they’d grown accustomed to using an AI-assisted system.

In rare-disease research, however, AI programs are making it easier for doctors to diagnose and treat patients with rare diseases, which you can often find in flashy magazine features and headlines. A New York magazine story published in January profiled a seriously ill infant with a rare genetic disease who was able to be treated effectively, thanks to an AI tool called BabyFORce, which looked at the gene causing the disease and whether its activity needed to be increased or decreased. “Some scientists posit that advances in AI platforms could lead to progress for all 10,000 [known rare diseases] at once,” the story reported. In a similar vein, the New Yorker asked last July “Can A.I. Find Cures for Untreatable Diseases—Using Drugs We Already Have?”

Most patients and health-care workers will never need to use life-saving AI tools such as BabyFORce. Instead, they experience the use of AI in more subtle, everyday ways, that are nevertheless completely reshaping the health-care experience. 

Allison Pugh, a sociologist at John Hopkins University and author of The Last Human Job, is concerned about what is lost when AI streamlines tasks: It saves time, but in the process cuts out human-to-human contact. In her work, Pugh coined the term “connective labor,” as a way to describe “the collaborative work of emotional recognition” that is so important in caregiving professions, such as health care and education, that AI can diminish. 

“Health care is so much more than downloading information,” she said. “If you want patients to hear you or to comply with treatment, they need to feel that you see them, and that’s an interactive process that practitioners get good at because they have to, under extraordinary circumstances. Health-care practitioners rely on connection with human beings, with their patients for a lot of the success of their efforts.”

“Without connective labor, health care will not work,” she avers. “Patients will suffer.”

A Nurse’s Perspective

Maggie Latona, a striking nurse at Mount Sinai Hospital, also worries about losing connective labor. She feels that Mount Sinai and the other hospitals involved in the strike (which continues as we go to press) won’t make strong commitments that AI will not be used to either replace her job or override nursing judgment.

“We’re not saying you can’t use some of these predictive models, but they’re saying ‘we can’t guarantee your job at the bedside, because we might be able to replace it with some sort of algorithm,’” Latona told The Indypendent. 

She says that AI is suited to some repetitive tasks, but “it is not going to be able to assuage [patients’] fears, right? It is not able to tell patients ‘These are things that we’ve actually seen happen, this is how we deal with them, and this is what our patients go through.’” 

Latona also fears that AI may be used to surveil nurses, even if that is not its purpose. Citing Mount Sinai’s “incredibly punitive culture around nurses,” and particularly the hospital’s firing of three labor and delivery nurses on the eve of the strike.

“If AI is able to track our charting, how much of its efforts are going to be spent tracking how long are we charting, why are we taking this long?” She said. “My fear is it would become ‘why are you spending so much time or so little time charting instead of being with the patient, or instead of making sure your documentation is correct and would hold up in court?'”

Latona and Pugh both worry that even if AI is able to cut down the amount of time spent on repetitive, administrative tasks, it may not necessarily increase time spent at patients’ bedsides. 

“I have a lot of sympathy for the health-care practitioners in this particular moment, because they are being compressed on both sides and then offered this technological solution,” said Pugh. She cited an AI tool called Scribe that summarizes patient-provider conversations, and thus reduces time spent on data entry. 

“I’ve talked to administrators, and they’re like, ‘Well, if we save an hour, I’ll let the physician or the health-care practitioner have 20 minutes of that, but I’m going to put more patients in there and take 40,” Pugh said. She believes that at this moment, practitioners are gaining more time from the tool, but predicts that will disappear within two years. 

“We do a lot of charting, and it’s very time-consuming,” Latona said. “But if AI is saving, say, an hour per shift on charting, then that only gives you another hour for another patient. It’s still pulling me in so many different directions, even if I’m not sitting in front of a computer. An extra hour doesn’t mean that I have the time to be able to safely care for yet another patient who needs far more time than what I would be spending at a computer.”

AI has not just stayed within the confines of hospitals or doctors’ offices. It has affected all kinds of medical fields. 

Dr. R, a psychiatrist who practices at a hospital in New York City and wished to speak anonymously, says he has serious concerns about how AI, particularly chatbots, is affecting people’s mental health. 

He limits his use of AI, to instances such as using a chatbot that only draws from evidence-based literature to assist with research. He is extremely wary of the idea that chatbots, particularly in their current form, could be used as mental-health treatment. He says they can, in fact, make people’s mental health worse. 

“Chatbots are not designed to be psychotherapists or doctors. They are designed to get people to use them and for those companies to make money,” Dr. R said when asked why he is against people using AI chatbots for therapy. “They are very nice,  but they don’t necessarily always give the accurate advice for the specific process that somebody might be using them for. And there’s, there’s a lot of evidence of people actually being led to have unhealthy dynamics with these chatbots.”

Dr. R pointed to the numerous stories that have exploded across media of people, even children, dying by suicide after using ChatGPT excessively or as a way to access mental health treatment. 

AI & Mental Health Care

AI has also slowly entered the market to meet the unsatisfied demand for mental-health providers. Dr. R understands that to an extent, but its use still concerns him. 

“There’s an access problem. There’s a cost problem, because some people don’t have insurance, or they don’t have enough insurance. And, there’s a numbers problem, which is that we don’t have enough providers to give people the amount of mental-health care that is needed out in the world,” he said.

“Clearly, if we could find a way to extend the care that we have, this sounds like a good thing, right?” he continued. “And I think it could be, but we need it to be done very carefully. Not as a replacement for human care for people who are struggling with mental-health issues, but as an extender.” 

Bias in AI is also a major problem. A Nature study found that “when prompted about speakers of African American English, large-language models spit out negative stereotypes similar to—or even worse than—attitudes held in the 1930s.” Another study, from Cell Reports Medicine, found that “AI used for health care is vulnerable to problems that have been observed in other fields, such as racial bias and gender bias.”

But it’s not like racism or sexism doesn’t already exist with human providers, Dr. R pointed out. “People who have dealt with inequities in our system don’t always have the best human-to-human experience in clinical settings. They may experience microaggressions or racism and these things, AI tools, if they’re done right and don’t show that same kind of bias, might be more comfortable for people.”

This observation is backed up by a study from the National Health Services in the United Kingdom, which found that nonbinary people and ethnic minorities who used an AI referral-assistant tool called Limbic were more likely to seek mental-health care. Dr. R thinks that’s because they felt less stigma when speaking to a chatbot than they would with a human provider. 

Regardless, he is “worried about profit-driven AI, because it certainly can keep things more efficient and cheaper, but could mean being implemented before it has been really shown, through research, that it actually meets the scientific standard.”

Ultimately, Dr. R said, he is “a firm believer in one of the core parts of human thriving is social connection.” 

“There’s no question that connection is at the basis of good medical care, of good psychiatric care. I can’t think of a way that AI can totally replace that. I don’t think that that’s happening now. But I think there could be threats of that.”

Dr. Christopher Kanyongolo, a Brooklyn dentist, was involved in a trial that looked at AI’s efficacy in diagnosing dental diseases. He was not very impressed with the technology. 

“It is currently very limited in the field of dentistry. It was over-diagnosing. Every little thing it perceived as pathology, which in some cases it wasn’t,” he said. 

Still, he said he would use it if the technology became more accurate. “AI is here to stay. It’s important in all other industries, like technology and business, and everybody’s using it. I don’t think health care is going to be an exception,” he said. 

Who Will Do The Diagnosis?

Dr. Kanyongolo emphasized that he would use it as a tool, but not as a replacement for his 20 years of experience.

“My worry is that it depends on how it was programmed: Who ended up putting these algorithms in place, or the designed the way these AI tools diagnose?” he said. “Who set up this technology? You have to use it as an aid to your diagnosis, not as the one doing the diagnosis.”

As negotiations continue during the nurses’ strike, Latona told The Indy that the New York State Nurses Association union is asking for a few basic commitments about AI in their workplace: that it doesn’t override clinical judgment, and nurses can escalate questioning it when needed; that they will not be disciplined for reporting errors in AI documentation; that it will not cause any diminishment of any in-person nurse/patient care; that management will give 30 days’ notice of changes and implementation of AI; and that management will allow nurse feedback after it is implemented.

“Nursing is such a unique profession because it is Caregiving 101,” Latona said. “You are with these patients, from their first breaths to their last, their hardest days and their most joyous days. Having a physical person there at the bedside, whether that’s for predicting care or being able to pick up subtleties that might not be present in something like vital signs, is invaluable.”

The Indypendent is a New York City-based newspaper and website. Our independent, grassroots journalism is made possible by readers like you. Please consider making a recurring or one-time donation today and help us reach our winter fund drive goal of $50,000.