In this discussion, speakers explore how the introduction of AI tools reshapes expectations in both healthcare and architecture/engineering (AEC). They argue that AI raises the perceived standard of care, leading patients to hold clinicians to near‑superhuman performance because doctors now have “the world’s knowledge at their fingertips.” This heightened expectation shifts liability from device manufacturers (e.g., MRI machines) toward physicians, who are expected to both recognize AI errors and follow AI recommendations when correct. The conversation highlights the paradox of humanizing AI—treating it as a decision‑maker rather than a simple tool—and how this perception influences clinician stress, burnout, and overall quality of care. Comparisons are drawn between AI and traditional medical devices to illustrate changing norms of error tolerance, and similar dynamics are noted in AEC where clients demand higher standards based on AI‑enhanced analyses.
Chapters
So with some colleagues, we've been looking at what will AI mean for the expectations that patients have of doctors, right? AI has sort of been promised to be this big burden lightener for physicians. You don't have to write your notes anymore. This is going to handle automatic prescription refills, all these things that should free up time, therefore increase your wellbeing, decrease burnout. But if you look at it from the psychological perspective, what AI also does is it raises the bar on outcomes, right? So, you know, if I'm a patient and I know my doctor's got access to a great AI-powered tool and they misdose my medication or they miss a diagnosis, I'm more likely to hold them to account because I'm thinking not, well, you know, every system is imperfect, everyone makes mistakes. I'm thinking you had access to the whole world's knowledge at your fingertips in a really easily digestible way. ChatGPT can tell me anything that I ask it, and yet you could have asked it and you didn't, or you asked it wrong and you got the wrong answer. So what we're seeing is physicians potentially actually feeling more stressed and more burnt out because now the expectation on them is so much higher, almost to these sort of superhuman standards that now they could never be wrong because they've got access to this great resource.
Do you use the term standard of care in healthcare?
Yes.
Yeah. So same thing in architecture and engineering. So it's raising the standard of care expectation from clients.
Correct.
The standard of care and then the willingness to tolerate error. Right. Which, you know, sounds terrible to say, but there is always error in any system and there's certainly error in the healthcare system. And so the more tools and resources we have, the more there is an expectation that we shouldn't make mistakes, that things should get caught, that things shouldn't get through the cracks. But what's very interesting to me is it, it comes down to, to how we're treating AI within the healthcare space. Is it a tool or is it sort of a decision maker? Right? Because ironically, if it's a, a tool, you might imagine that, so, so AI hallucinates, it comes up with the wrong answer, it advises the doctor incorrectly, right? That's in some ways no different than the MRI machine at the hospital being miscalibrated and not showing a mass when there was one, right? If the MRI machine is miscalibrated, people aren't gonna sue the doctor.
They, and sorry, and there was an error or something bad happened, right?
As a patient, I'm not pinning it all on the doctor.
I'm thinking I'm going after the hospital or maybe the manufacturer of the MRI machine.
But with AI, what we're seeing is that people go, well, the physician should know not to believe the AI output when it's wrong. But the physician should also know to follow the AI output when it's right. And we're essentially ascribing the blame to the doctor in the same way we would if that doctor had asked like their knucklehead colleague who hallucinates from time to time, like, oh, you asked Chris and Chris gave you a really terrible suggestion. You should have known not to listen to him. Right.
Right.
But this time Chris gave you actually a good answer and you should have listened to Chris.
Right. And so it's quite odd, in fact, that we're humanizing AI in a way when we draw these expectations around what we want physicians to be able to do with it, right? Where we're not treating it as a tool where if the tool's wrong, the individual user's off the hook.
'Cause how would they have known that the MRI machine was broken or that the AI gave them a decision made off of made-up studies, right? That happens all the time with these predictive algorithms, right?
So how we think about it, how we perceive AI actually has a huge impact on all the outcomes that we care about in healthcare, our ability to deliver high-quality care, the wellbeing and stress and burnout of the people who are delivering care. So I've been fascinated less with the nuts and bolts of is it working well, what do we do with it, what can or can't it help us with, and more just how are we even thinking about this?
As a, a new tool, because that's one area where it does feel kind of different than computers or the internet is, you know, when we got digital CAT scans and things like that, I, I don't think anybody then expected doctors to perfectly calibrate and know exactly when to trust that and exactly when it might be having an error. We, we sort of held the manufacturers to a standard to say you need to manufacture it within these tolerances. I'm seeing a different conversation happen around AI where much more of the burden is being shifted to the sort of end user, in this case the physician, to know, all right, is this useful information or not?
And so it feels like it's a, it's a very different perception there.
