If you don’t work in healthcare, then you may not know that radiologists are known as “hedgers.” That is, radiologists will say things in ways that avoid being definitive on anything. Maybe this is pneumonia. Maybe that is a mass. “No definitive fracture is seen”—does that mean a possible fracture is seen? Admittedly, radiologists can use this crutch too often, but there are also some really good reasons to not be completely certain. For example, sometimes the science remains unclear, and no one knows the answer yet. I remember one older radiologist telling me one way to deal with this type of uncertainty: “Just let them know that you are sure you don’t know.” Instead of saying that a finding “may” be related to a chromosomal abnormality, for example, say very clearly: “This finding has been associated with a chromosomal abnormality, but the literature remains uncertain about the connection.” At least the doctor reading your note knows that you aren’t just being vague—the science itself is not clear.
In radiology, your “customer” is typically another doctor, but for most other doctors, your customers are your patients. Patients, of course, want certainty, but that’s not always possible. Instead, what we can offer—and what they truly crave—is confidence.
This confidence comes in at least three flavors. The first is confidence in your knowledge. You’ve just given a diagnosis, for example, and the patient wants to believe that you got it right. This is where certainty comes in, but certainty doesn’t come for free. It has to be grounded in two sources:
- Deserved certainty: You’re trained. You know the science. You have the right to speak with authority because of your education and experience.
- Earned certainty: You’ve listened. You understand their unique case. You’ve considered alternative possibilities. It’s not just about being an expert, it’s about showing the patient you’ve applied your knowledge to them, specifically.
This is the area where I think a lot of doctors fall short. We rely on our credentials and assume that’s enough. But patients want to feel that we understand and are reasoning through their particular situation. Ironically, that’s something that non-experts sometimes do better. They listen closely, express empathy, and create a sense of personal investment. Even if their expertise is shaky or nonexistent, the confidence they inspire can feel stronger than that of a rushed, distant expert.
Certainty in a diagnosis is nice, but we don’t always have the luxury of knowing for sure what’s going on. The diagnostic process can take time, and in cases like that, we need to project the second flavor of confidence, confidence in the process. You don’t have the answer yet, but what you do have is a plan: These are the next steps we’re going to take to figure things out, and in the meantime, here’s how we’re going to address the symptoms. Patients know we don’t always have an immediate answer. At a minimum, they want to know what happens next. They want to know that someone has a firm grasp on how to navigate the unknown future.
Medicine is a complicated field, however, and sometimes things are so confusing, we aren’t even sure what the next step should be. Maybe we need more time to think, or maybe we need to watch how the condition evolves, so we have more information. Now we have to rely on the third flavor, which is confidence in the relationship. When there’s no clear answer and no clear next step, what matters most is that the patient knows that someone will be there to take care of them. The unknown is a scary place, but it’s even worse if the patient is alone or abandoned.
This is part of the “art of medicine” that people talk about. The part you don’t learn from a textbook or from training, but you learn along the way. This is the part that is becoming even harder to practice, as healthcare becomes more fragmented, with less continuity of care, shorter appointments, and a reliance on less personal forms of communication, such as messaging and video.
This brings us to AI (you knew I was getting there, didn’t you?). One of the challenges with the current state of AI, is that our large language models are always certain. They are designed to create both plausible- and confident-sounding language, so that’s what they do. If you challenge them, they might dig in to their original stance, or they might immediately capitulate and tell you that you are right. But neither of these is necessarily based on what is true, so we don’t know what we can trust, or when.
We’re going to have to tackle this certainty problem as we learn to use AI; it sounds like a person, but it’s not. And our instincts, honed over millennia to gauge when to trust another human, are going to fail us. At the same time, those of us who develop AI tools are going to have to find a way to get these systems to express certainty or uncertainty, as the situation requires. A system that is often right, but confidently and opaquely wrong some of the time is not going to get us very far.
Maybe the most important thing that AI is doing right now isn’t about what it gets right or wrong, but is about showing us the cracks in our own systems. When AI applications can pass a medical licensing examination, that may not mean that AI can practice medicine, but it may mean we aren’t that good at assessing medical skill. And when students use AI to write their papers for them, the answer may not be to create AI detectors, but to ask ourselves what we’ve done to instill in our kids a desire to complete the assignment for a grade, rather than to actually learn.
If people are turning to AI to confidently (and opaquely) explain their diagnoses to them, we can complain about it or call for regulations. But we should also ask ourselves if people aren’t just trying to find the confidence they crave, wherever they can. And we should ask why, if we’re the experts, so many people still look elsewhere for the trust we think we’ve earned.
Confidence isn’t just what we project, it’s what others receive. And if we don’t give people a reason to trust us, someone—or something—else will.
If you know someone who might like this post, please feel free to share it with them. If you’d like to receive these posts (and a few bonus posts) by email, subscribe here.
Red Goggles 