When I wrote my first novel last year, I grappled with the ethical implications of using a large language model to guide my writing process (one I have been carefully refining for decades). Prompt after prompt, my mind became inundated with the notions of how every query consumed a not insignificant amount of energy and the black box of the generated text’s origins. “Am I relying too heavily on this new technology? When does this cease to be my work? Where is the line, and have I crossed it?”
I quickly realized that GPT-4 was not the creative writing magician many had touted it to be, capable of instantly replacing the human mind with its impeccable artificial intelligence. Was it good? For some things, yes. It could construct believable dialogue and move between well-described scenes with ease. In fact, it was apparent how it could synthesize the facts needed to build compelling arguments in an essay. However, its storytelling abilities fell apart when it came to some of the creative writer’s most powerful techniques: thematic consistency and nuance.
For the chatbot to produce something that fit within the world I built, I had to guide it, sometimes to extraordinary lengths. For every 800 words it cobbled together, I had to feed it at least half of those to keep it on track. I wadded up many of its suggestions like scrawled-on paper and tossed them into my laptop’s trash bin. I heavily edited other ones and ultimately used them. Although it was far from a fully automated experience, using AI helped me quickly find the path to the right ideas, allowing me to write efficiently and complete my book in under two months. Holding the first copy of my book in my hands, which took over nine years to come to fruition, brought tears to my eyes and an indelible feeling of accomplishment.
Whether we like it or not, AI (just like televisions, computers, the Internet, and mobile phones before it) has rapidly permeated everything we do. Although it has been like opening Pandora’s box with no sense of whether AI is the evil that escaped or the hope that remains within, AI is most assuredly a permanent fixture in our lives now. I have been leery of the human-fueled explosion of systematically untested AI products and services because, as a species, we have not yet established the ethical and moral guardrails for this technology, such as Isaac Asimov’s Laws of Robotics. “Where is the line, and have we already crossed it?”
For my day job, I work as a family physician, caring for patients in a federally qualified health center. As a young Millennial, I grew up during the inexorable technological changes of the late 1990s and 2000s. By all accounts, I graduated from high school with my diploma in techno-literacy. After four years of college, one year of graduate school, and four years of medical school during the 2010s and early 2020s, I began my residency training to become a primary care physician with a firm grasp of the skills needed to tame the tiger that is the electronic health record.
Even though I had been raised to master the intricacies of these tools designed with medical billing in mind first and foremost, I struggled with the same challenges that more seasoned clinicians had articulated years ago. If I ditched my computer for more attentive interactions with my patients, I would inevitably be strapped with many hours of charting that I would have to do without pay after work, which would catapult me toward burnout and a less productive, and likely shorter, career in medicine. So, I did what most have felt forced to do: look at the computer screen in the patient’s room to chip away at the mountain of required documentation while trying to provide high-value health care.
Over the years, I have made this computer-doctor hybrid look natural and function as well as possible, with admittedly variable results. When I started hearing about the new applications of various AI models that utilize ambient listening to document clinical encounters like an automated medical scribe, I was intrigued. Excited, even. It reminded me of two things: how the chatbot tapped into my deeper potential as a writer and a prescient conversation I had with an emergency physician about this very concept when I worked as a medical scribe.
With many of the available products, I could download an app, press a button, and allow the AI to record the conversation, transcribe it, and summarize the pertinent clinical information into a more or less complete note. Once I started seeing ones built into electronic health records and other vetted, HIPAA-compliant resources like Doximity, my ethical concerns about the technology waned. After being trained to use one of the so-called “AI scribes” at my current workplace, I was eager to begin using it and regain at least a small piece of the professional autonomy that electronic health records had taken away from physicians long ago.
When I entered the room of my first patient of the morning and obtained his permission to use the tool, I nervously clutched onto my scrap paper and clipboard as I launched the mobile version of the AI scribe. Thereafter, the gentleman and I had a productive conversation about his health care needs, one that went much deeper than what would have been possible with a screen obstructing my face. After pressing stop, I anxiously returned to my desktop to see what the scribe managed to put together in about 60 seconds. A giant grin reached from ear to ear as I saw a complete history of present illness, review of systems, physical exam, and a mostly complete assessment and plan in the system.
Since incorporating the AI scribe into my practice, I have been happier as an outpatient family physician, connecting with my patients and coworkers in new and gratifying ways. Although the technology is new and not yet perfect, it allows me to go home with all of my notes locked and focus on my wife and pets, which is priceless. What I experience when I use the AI scribe is what I felt when I used the chatbot to write my book: present-mindedness. I can just be present in the moment, which is conducive to imagining more immersive stories, having more meaningful conversations, and flourishing in the day-to-day grind.
Even in my gratitude toward being able to be more present with patients, the black box of AI’s environmental impacts is not lost on me. Last year, a group of family medicine resident physicians and I wrote about the environmental hazards associated with the mass storage of data in electronic medical records. This problem is amplified when we incorporate AI scribes and other AI technologies into the mix, which requires much more energy to generate results than simple data retrieval from a human analyst. Although these technologies are providing objective benefits to clinicians right now, we cannot bury our heads in the sand and pretend that there are no significant tradeoffs. Ethically, we have a responsibility to consider all the pros and cons of any action we take in health care.
As a global community, we face numerous critical ethical questions regarding the existing and emerging uses of AI technologies in all aspects of our lives. While automating everything humans currently do with AI is dystopian and existentially paralyzing, there is something beautiful and humanizing about the enhanced experiences we can have with AI’s support when we explore our creativity and deliver care to others. Suppose we can draw a clear line in the sand regarding the utilization of AI that prioritizes the bioethical principles of beneficence, non-maleficence, autonomy, and justice. In that case, we may be able to enter an era where we live happier lives and get health care from professionals who are also happier with the work they do.
Ashten Duncan is a family physician.
