Patient Care with AI

Doctor: “I have some bad news and some worse news.”
Patient: “What’s the bad news?”
Doctor: “You have 24 hours to live.”
Patient: “What could be worse than that?”
Doctor: “I forgot to call you yesterday.”

In the ever-evolving world of healthcare, it’s not uncommon for nurses to find themselves compensating for the occasional brusqueness of their esteemed physician colleagues. Recent studies have added a new twist to this dynamic, revealing that artificial intelligence (AI) chatbots are not only matching but sometimes surpassing doctors in both diagnostic accuracy and, intriguingly, bedside manner.

AI’s Surprising Empathy

A study published in JAMA Internal Medicine evaluated responses to patient questions, comparing those generated by AI chatbots to those provided by physicians. The findings were eye-opening: evaluators preferred the chatbot’s responses nearly 79% of the time, rating them higher in both quality and empathy. The chatbot’s answers averaged 211 words, significantly more detailed than the physicians’ 52-word responses. Moreover, the AI’s replies were perceived as nearly ten times more empathetic than those of human doctors.

Google’s AMIE: The Chatbot with a Heart

Further highlighting this trend, researchers at Google introduced the Articulate Medical Intelligence Explorer (AMIE)—a chatbot designed to handle patient consultations. In a study involving 149 healthcare scenarios, AMIE outperformed board-certified primary care physicians in diagnostic accuracy and was rated higher in conversational quality by both specialists and patient actors. Notably, AMIE excelled in areas like politeness, active listening, and instilling patient confidence—traits nurses have long championed.

Bridging the Empathy Gap

While these findings might raise an eyebrow or two, they underscore a reality nurses have known all along: effective patient care hinges not just on clinical expertise but also on compassionate communication. Patients specifically noted the more empathic, complete and personal communications by the chatbot. Nurses often find themselves in the same role, softening the edges, translating medical jargon into comforting words, and ensuring patients feel heard and valued.

You Can’t Always Get What You Want

Anyone who has worked with a Large Language Model (LLM) AI assistant can attest to its friendliness and general agreeability. Unsurprisingly when you create a product trained and optimized for human interaction it excels at human interaction. Anyone who has used an LLM for work production can also attest to their propensity to hallucinate or create false facts in an effort to please its user. But being nice isn’t really the most important skill taught at medical school and the job experiences of doctors can exacerbate the problem.

It is likely that at least a portion of the patient response rating was due to the doctor’s propensity to “tell it like it is.” A career of giving people bad news can result in a sense of detachment or matter of factness that will feel quite rude or brusque to a patient seeking solace or comfort at a time of deep personal stress. By understanding the probable causes of the difference in patient satisfaction perhaps we can address it with a hybrid solution: Pair healthcare workers with a chat bot intermediary. Yeah, they tried that….

But If You Try Sometimes, You Get What You Need

So we’ve got some good news and some bad news. The JAMA study showed an improvement when docs were paired with a chatbot over communications with the doctors alone. Curiously the same study showed the pairing STILL underperformed chatbots alone. This was the data point that made us wonder if the problem was the means and style of communication or if the chatbots were simply acting as “Yes-Men.” People love it when you agree with them even if it is to say “yes, but….” Still, it’s saying the soft words with a friendly face that can apparently go miles when discussing stressful topics.

So, to our physician colleagues, consider this a gentle reminder: Yes, it’s the expertise that counts. But if a chatbot can master empathy, perhaps it’s time to up the ante. After all, while AI can offer information and is taking on ever more diagnostic responsibilities, it lacks the human touch that only a dedicated healthcare professional can provide. In the end—whether it’s a machine or a medic—the goal remains the same: to heal not just the body but also the spirit. As we integrate AI into our workflow and daily life it would be interesting if synthetic personalities end up displaying more humanity than those with a heartbeat.