incorrect diagnosis, it's

often, as you said, because they leak to the answer.

They think intuitively, and instead of considering seven or eight things, which are often all possible diagnoses, they'll just jump to the first one.

And we've done a lot of work, as have many others around the country, on diagnostic error.

And much of the early work was done in the outpatient setting.

There, a lot of the errors were things like failure to follow up on a test that you did or that sort of thing.

Turns out there are lots of diagnostic errors in the inpatient setting too.

It might make it harder for doctors to have those tough conversations if they're used to something always agreeing with them at home.

David Bates from the Department of Rhetoric at the University of California, Berkeley.

Great.

Thank you so much.

This has been great so far and glad to be a part of it.

So I want to start really with a basic kind of principle, which is that humans are defined by their exteriorization of memory.

As Bernard Stigler argues, it's technical from the very beginning.

It first takes shape as a stone tool millions of years ago and then transforms itself into memory.

of the digital revolution.

What about, and here I'm gonna get into value-based purchasing, or at least the, the, the foundations of it, what about measurement itself and making measurement real time? What can be done there? What work has to be done to make sort of a measurement system that uses these tools and is outcomes-oriented?

It can be as much as, say, 20 times as fast, so it can be r- really a great deal faster than our old processes, and it's particularly good at, at measuring things even in real time.

So if you, you know, tell it to do the right thing, you can have it basically looking over the shoulder of the physician and, uh, producing a summary of things.

AI will change the whole way that workflow works for many conditions.

So I've done some work with some groups in Scotland who are looking at stroke care, and the way things work now in stroke care is patients in Scotland are distributed across the country, but they'll go to a small place with a suspected stroke.

If they have that, they'll get imaged rapidly.

The first thing that this information goes to is the AI, and the AI, you know, has a look and says, "Well, this looks like a bleed, it looks like an infarct, or it doesn't look like a stroke at all." And then that information goes to the, uh, clinicians, and there's both a, a neurologist and an emergency room physician.

What, what kind of programs are you a- applying or experimenting with at the Brigham?