Research indexRCT
Artificial Intelligence
Newitt
Oncology Times27 Mar 2026
NobleIDark:/48914/w1/20261105/4D15BF7ADOI10.1097/01.cot.0000000000000333
Artificial Intelligence
Newitt · Oncology Times · 2026 · Subscription
NobleIDark:/48914/w1/20261105/4D15BF7ADOI10.1097/01.cot.0000000000000333
The full text is at the publisher
You can read the abstract here, or open it at the publisher.
Abstract
Is artificial intelligence (AI) a superpower for oncologists, rendering them capable of delivering faster diagnoses, streamlined workflows, and fewer errors? Or does this new-world order of key words and computer algorithms hinder the development of stellar healthcare professionals? The answer to both questions could be a qualified “yes.” So, it's not surprising that AI has been described as “a double-edged sword.” “I'm an advocate for a critical assessment of the use of AI,” said Edward C. Halperin, MD, MA. “Some say, ‘You can't argue with the tide and tell it not to come in.’ I assert, instead, that you can and must direct the tide's flow.” Halperin is Chancellor/CEO and Professor of Radiation Oncology, Pediatrics, and History at New York Medical College, and Senior Editor of Perez, Brady, Halperin, and Wazer's Principles and Practice of Radiation Oncology, the best-selling textbook on radiation oncology in the world. While he believes AI should be rigorously tested to ascertain if it has a place in the medical toolkit, he also stresses it lacks the most important characteristics of quality healthcare, characteristics that are, in the simplest terms, human.Edward C. Halperin, MD, MAThis practicing clinician and educator is optimistic that AI will successfully assist physicians and patients in some ways. Digital pathology may prove to be successful in scanning slides and detecting abnormal cells that may not be easily seen by the human eye. “Some highly repetitive tasks in medicine carry risks of fatigue. Just imagine a cytotechnologist looking at the recommended maximum of 40 to 100 Pap smears a day. Humans get tired, computers do not. There's a promise that scanning technology could look at a smear and flag something questionable,” said Halperin. “Or let's say I type a radiation therapy prescription into the computer for the electronic medical record (EMR), ordering 60 Gy of fractionated external beam radiotherapy in 6 weeks to an area of treatment. A message might pop up on the computer screen and say, ‘Are you sure you want to order that dose? This is an unusual dose for the tumor stage, histology, and anatomical location of treatment that you entered earlier in this medical record.’ These are potential uses of AI in cancer care that could help reduce errors and improve clinical outcomes.” Halperin's concerns do not rest on algorithms or scanning software, but with something more basic and fundamental: how doctors learn. Lost Art of Browsing The downside to AI technology may be its negative influence on human learning/probing, critical thinking, listening skills, and verbal and written communication. For Halperin, it begins with the declining use of textbooks. “Some medical and dental students proudly post on social media that they have gotten through 4 years of school and never bought a textbook, relying entirely on what they find on the internet,” he said. As a result, Halperin has seen library space allotted for medical textbooks and journals decrease. Additionally, browsing in bookstores has been replaced with online purchases. What's wrong with that? “We acquire knowledge by browsing,” he said, recalling roaming through a library as a student, searching for a particular book, then discovering an adjacent shelf of intriguing volumes. “I flipped through them to see what was interesting and picked up much more than just one book on my topic.” And while books offer tables of contents that might spur investigation into an additional branch of interest in a topic area, an internet search delivers information on only the original search topic. You type in a question, and you get a response. But such efficiency is delivered at a cost. Intellectual wandering disappears. “If you type in, ‘What is the dose of cyclophosphamide for a certain condition?’ you get back an answer and you're done. You don't read why. You don't read about pharmacology. You just get a number,” explained Halperin. Similarly, oncology textbooks converted to searchable eBooks stunt exploration. “You look up staging for cervical cancer, for example, you get the staging table, and you move on. You don't roam around the chapter. It's all there for the taking, but you're done,” said Halperin. Perhaps most concerning in the information-gathering arena is student dependence on the internet and social media, which Halperin called “unpoliced bulletin boards in the corridor.” He stressed, “Anyone can post anything. In contrast, a well-written oncology textbook is the equivalent of having a skilled guide accompanying you on a review of the essential elements on an oncologic topic, pointing out what is and is not crucial, synthesizing information, summarizing key concepts, and directing the reader to the most important references.” Erosion of Attention How do all the above impact practice? Halperin said a concerning professional erosion is visible in the clinic as the attention span of healthcare providers interacting with patients grows ever shorter. “What are patients fundamentally unhappy about with healthcare? They say, ‘Nobody listens to me.’ But clinicians have become accustomed to information in tweet-like lengths,” he said, noting that studies indicate that American physicians typically interrupt a patient within 11 to 18 seconds of that patient beginning to speak. “They're in a rush to get to closure.” AI accelerates these tendencies, Halperin worries, because it rewards speed over reflection, answers over insight. AI also generates plausible answers without regard to ethics that a human answer might convey. “AI operates without human experience and judgement,” he said, citing cases where chatbots provided harmful advice to users. “You might type in, ‘I don't see any purpose in life and I want to end it all.’ The response might be, ‘Here are some options for committing suicide.’ Or, ‘I'm not getting along with my spouse. What do you recommend?’ And the response might say, ‘Have you considered a hit man?’ There are multiple lawsuits about adverse outcomes where AI may have contributed to harming individuals.” Electronic Medical Records (EMRs), though indispensable, have made the clinical “support” provided by computers downright cumbersome, negating some of the time savings it has purported to capture. Endless notes with copied-and-pasted lab values, imaging reports, and prior documentations have become unwieldy. “We have an ideal that doctors look at you, listen to you, care about you. But the reality is that doctors very often have their backs to you, typing into the computer, and they want to get out the door because they have three more patients to see at this hour to satisfy the boss. Then they end up reading page after page after page in the EMR before ever finding what happened in this visit today,” said Halperin. The clinician's task has shifted, at least in part, from critical thinking to documentation and time management. AI Education “It is the responsibility of medical schools and residency programs to address AI concerns and educate around the issues,” asserted Halperin. To that end, he commented on a few concrete steps that can help in general medical and specialty oncology education. Subject AI to strict scrutiny: “AI tools should be subject to randomized prospective clinical trials to ascertain safety and efficacy. We must guard against two things. First, the risk that commercial interests will try to avoid strict scrutiny by claiming their products are ‘just an extension of existing technology.’ Second, that manufacturers will be allowed by a neglectful government to write the oversight regulations which is, assuredly, the fox being tasked with guarding the henhouse,” warned Halperin. Proctored examinations: “Yes, it is harder to run a proctored exam for medical students than a take-the-exam-on-your-own-on-the-honor-system test. Yes, it is harder to grade an essay examination than a multiple-choice exam. Yes, it is harder to administer and grade an oral examination. But students can't use AI to write answers on proctored essay exams or to generate a sustained idea on an oral exam,” said Halperin. “Learners will respond to the desire to get good grades by engaging in behaviors that will improve their chances of doing so. Faced with essays and/or oral examinations they will perforce reading and think about what they have read and invest time in improving their writing skills. Some medical specialties, like radiation oncology, still have oral board exams. Others should add them.” Adopt a camera-ready culture in online education: “Forbid trainees to attend online classes with their cameras turned off. This is an invitation to attending educational sessions in pajamas while scrolling on a cell phone or asking AI to give the answer to a question just posed,” said Halperin. “Any physician who participates in a telemedicine visit with a patient with their camera turned off and inappropriately dressed is engaged in unprofessional behavior. We should model proper behavior in our online clinical conferences, teaching rounds, and tumor boards.” Foster education activities requiring individual engagement with a particular text: “Consider a journal club in which a recent important article published in peer-reviewed literature is discussed, or a book club in which one or more chapters of a book are discussed,” suggested Halperin. The Big Picture Halperin is one of many educators who are sounding a warning call about the lure of AI shortcuts. “It is being discussed at medical education meetings. Some people might think I am overly cautious, but I believe we need to critically analyze the use of AI. We should speak up when we note potential downsides associated with technology,” he insisted. The real danger, as he explains it, lies in AI replacing human judgement rather than supporting it. “We must teach medical students to think critically for themselves and evaluate information—including critically evaluating what AI does. I hope there are enough people willing to do that.” AI can take a responsible place in oncology. But physicians must have the courage to resist so-called efficiency over excellence. The future of cancer care, suggested Halperin, may depend less on smart machines and more on clinicians who remain fully present, curious and, well, just human. Valerie Neff Newitt is a contributing writer.