What if AI could have your difficult vaccine conversations for you? No more contentious interactions with a neighbor or frustration from addressing the same viral claim for the 100th time with a parent. The moment you felt disagreement brewing, you could pass it off to the AI agent. Would you take that offer?
I would not.
Hard conversations about topics like vaccines are a part of our infuriatingly beautiful human experience. And as AI advances, we will have to choose whether to continue leaning into those uncomfortable conversations or offload them to microchips, giving up a piece of our humanity in the process.
I recently saw a newborn in my clinic. I know the family well and have cared for their other kids. So, I was confused when I couldn’t find a hepatitis B vaccine dose recorded in the nursery birth records.
“Did Sam get his hepatitis B vaccine in the hospital?” I asked.
The mother hesitated before replying, “We know you recommend vaccines, but we don’t want him to get any.”
I was genuinely surprised. I knew they had vaccinated their older children because I was the one who had vaccinated them. In fact, just several months ago, they had vaccinated one of their older children. They trusted me before, and now they don’t.
I felt confused, angry, and sad. I was sad at the possibility of their newborn not getting the protection I know could save their life. I was angry at social media, our health leaders, and everyone who profits from parents’ fear. And I was confused at the sudden reversal. It had already been a long day, I was frustrated, and it was harder than usual for me to respond with empathy.
You know who doesn’t get angry, sad, or frustrated? You know who has endless empathy (albeit completely simulated and deceptive)? AI. It’s also getting good— scary good.
Many people are skeptical of AI’s abilities to the point of delusion. In reality, AI is evolving faster than the fields of vaccine research and health communication research can keep up with. It’s not only becoming more adept at providing clear, accurate information, but it’s also quickly becoming good at mimicking emotional language and even demonstrating “empathy”.
Over the past several years, I have had conversations with AI where I pretended to be an anxious parent who is hesitant to vaccinate my child. Even a year ago, the dialogue was robotic, distant, and entirely unapproachable. It now explores my hesitancy with open-ended questions, affirms my concerns, supports my autonomy, and makes a recommendation, all while providing clear evidence. It’s like it read the communication textbook (it did), and it will only get better.
I could have responded to Sam’s family with, “Tell me why you don’t want to vaccinate, and I will systematically debunk claims you have heard with my expertise.” For many reasons, I didn’t say that. Honestly, this is where AI would beat me easily.
When it comes to specific knowledge, AI is more of an expert on vaccines than clinicians. Try it for yourself. Tell your favorite AI that you are worried about the amount of aluminum your kids are getting in vaccines and ask it to explain the risks and benefits. It will walk you through the evidence, distilled to digestible bites, and contextualize the information. Will it hallucinate a reference or get a piece of information wrong? Of course, but clinicians are also far from perfect in their knowledge, and clinicians or parents doing their own searching and research isn’t without risk.
But AI is doing more than just providing facts like an internet search engine on steroids; people are looking to it because they feel heard. Of course, it cannot really understand or empathize without experiencing what it means to be human. This “deceptive empathy” creates the illusion of connection and reassurance, tempting a shortcut around the hard work of real, lasting human connection.
Sam’s family had access to me, their pediatrician. They have health insurance coverage and were able to get a timely appointment.
People are also turning to AI because it is accessible and fills a gaping hole in our health systems. Finding timely, responsive, and affordable answers to common questions and concerns within our health systems is elusive for far too many people. If you are a parent with vaccine concerns who is fortunate to have a primary care provider for your child, you can schedule an appointment in 3 months, send a chart message and get a truncated, hurried response, or pull out your phone and get personalized answers from a seemingly super-genius expert within seconds.
At least 1 in 3 adults report using AI for health advice, and that is precisely why we must be talking about this. Simply dismissing AI the same way people should never look for health information on the “net” or “World Wide Web” (hello, 1990s) is not an option because people are using it. And there are serious potential harms to AI use for health advice, like inaccurate diagnoses, inconsistent information, or failure to recommend urgent evaluation, which need to be mitigated.
After taking a deep breath, I asked Sam’s family, “I’m curious: you trusted me when I recommended vaccines for your other kids; what made you change your mind now?”
His mom started crying and said they were feeling confused and overwhelmed. They didn’t know who or what to trust, and they had seen so many things that made them question. They felt lost.
At this point in the conversation, AI could have given them directions. It could have analyzed their questions, responded accurately, drawing from a much deeper pool of science and evidence than my brain has access to, and laced its response with deceptive empathy. It could probably do a better job of addressing the claims they had been exposed to, giving the facts, and reasoning with them than I could.
But would they have trusted it if it had recommended they vaccinate?
Ironically, medicine, science, and public health often mistake human beings for logic machines in health communication and decision-making. But that’s backward; that’s AI, not humans. Humans make decisions based on much more than logic: values, emotions, and experiences. If humans were logic machines, then a super-advanced AI would be the ultimate solution to vaccine communication.
I responded to Sam’s family, “I understand why you are feeling confused. There’s a lot of information out there. Ultimately, this is your decision to make, but just like your other children, I believe these vaccines are important, and I recommend he get them.”
At some point in every conversation about vaccines (or any health decision), there is an implicit ask for trust. Even if small, there is always a leap of faith. Medicine doesn’t deal in certainties; it deals in probabilities. No matter how strong the weight of the evidence is, at some point, I reach out my hand across a divide and ask a family to trust me. Not because I have the most knowledge or sound reasoning, but because I love my patients, have their best interests at heart, can empathize with their human experience, and have shown up for them, time and again, even when it’s hard. AI can’t say that.
I am assuming that the alternative to AI is an available, skilled, and empathetic clinician. As I have said, many families do not have convenient access to that, and as a society, we must work to decrease those barriers. AI can be upscaled to reach millions, whereas I can only talk with several dozen people each week. There is massive public health potential there that shouldn’t be ignored. Is AI better than a parent searching social media, asking a Facebook group, or asking neighbors? Probably. But reach isn’t the same as depth. I find myself disengaging more and more from social media as it is flooded with AI content (mostly slop). I crave embodied human interactions. The human slop on social media is not good, but at least it is human.
Sam is now fully up to date on his vaccines.
Not every story ends this way, but this one did. I cannot help but think that an AI, without the relationship and trust, could ever have walked with the family to get to that point. Even if it could have, I wouldn’t have wanted it to because I am also more human for having leaned into the hard conversation. And in the process, I gave them one of the most beautiful and important gifts one person can give another: my attention, time, and care. If our communication is reduced to only agreeable, information-sharing pleasantries, we become no different than the AI tools themselves.
We have agency with this decision. Abdicating our responsibility to engage with each other on difficult topics means giving up part of what makes us human, and I, for one, will choose to keep engaging.
-David
Note: As with any piece on AI, this post probably won’t age well. I hope it’s wrong. I hope it’s challenged. I hope you disagree. And I hope we have difficult conversations about it. Because that means it’s human.
*The name and specifics of the patient and family I cared for in this article were altered to protect their privacy.
**AI was not used in the creation of this writing.
Community Immunity is written by Dr. David Higgins, MD, MPH, a practicing pediatrician and public health physician whose work focuses on vaccine delivery, health policy, and communication. This newsletter is where he writes about vaccines, public health, and community. When he’s not seeing patients or writing, he’s coaching youth soccer or exploring the Colorado outdoors with his family. Find him on LinkedIn, Instagram, Substack Notes, and Bluesky.
This newsletter is free for everyone. If you want to support the work that goes into it, a paid subscription means a lot.
No posts
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.