AI Rounds by the Cumming School of Medicine
Sep 9, 2026 · 21 min · 11 segments
If a learner rehearses a difficult conversation with an AI, what exactly are they practising against? A listener asked, and the answer runs through two layers of embedded value: the persona an…
A listener wrote in after listening to episode eight, learning on your own terms, and they had some thoughts.
In this article, the magazine ran 25 frontier AI models through the World Values Survey, the instrument that has mapped human moral belief across more than 100 countries since 1981.
And they found these models hold values quite unlike those of most people alive.
If the tools carry documented value biases, then what are our learners rehearsing against? That's a good question, and it deserves more than the 30 seconds I gave it in episode 8, so I'd like to give it the moment it deserves now.
The student described using AI to rehearse difficult conversations, and she was accurate about what she was doing and why it was helpful.
But what neither of us discussed was the second half of the question, not whether this works, but what exactly a learner is practicing against when they do it.
That's a live question and an unsettled evidence base, and it's worth 20 minutes on its own.
But what I don't want to do today is spend that 20 minutes on a student's study habits, because the phenomenon is really much bigger than that.
So what our student described as an improvised study habit has actually become an industry.
There are now purpose-built simulated patient platforms across medical, nursing, physician assistant, and pharmacy education.
Systems for serious illness communication in oncology, systems for diagnostic reasoning, virtual patient simulators tested head-to-head against actor-based simulations in randomized crossover trials.
Human standardized patients are expensive, scheduled, finite, and in short supply.
A tool offering unlimited reps at near zero marginal costs solves a problem that students genuinely have.
It's a question about a category of educational technology that many of you will be asked to evaluate, purchase, or defend in a committee meeting at some time in the near future, which makes it worth understanding where a simulated patient's values come from.
A listener wrote in after listening to episode eight, learning on your own terms, and they had some thoughts.
In this article, the magazine ran 25 frontier AI models through the World Values Survey, the instrument that has mapped human moral belief across more than 100 countries since 1981.
And they found these models hold values quite unlike those of most people alive.
If the tools carry documented value biases, then what are our learners rehearsing against? That's a good question, and it deserves more than the 30 seconds I gave it in episode 8, so I'd like to give it the moment it deserves now.
The student described using AI to rehearse difficult conversations, and she was accurate about what she was doing and why it was helpful.
But what neither of us discussed was the second half of the question, not whether this works, but what exactly a learner is practicing against when they do it.
That's a live question and an unsettled evidence base, and it's worth 20 minutes on its own.
But what I don't want to do today is spend that 20 minutes on a student's study habits, because the phenomenon is really much bigger than that.
So what our student described as an improvised study habit has actually become an industry.
There are now purpose-built simulated patient platforms across medical, nursing, physician assistant, and pharmacy education.
Systems for serious illness communication in oncology, systems for diagnostic reasoning, virtual patient simulators tested head-to-head against actor-based simulations in randomized crossover trials.
Human standardized patients are expensive, scheduled, finite, and in short supply.
A tool offering unlimited reps at near zero marginal costs solves a problem that students genuinely have.
It's a question about a category of educational technology that many of you will be asked to evaluate, purchase, or defend in a committee meeting at some time in the near future, which makes it worth understanding where a simulated patient's values come from.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.
