Dr. Teodor Grantcharov, a surgery professor at Stanford, has been developing an “operating room black box” for 20 years.
I wrote about how much AI has advanced in 21 months, with examples from video to music to text, as well as AI adoption.
Also, otters on airplanes using wifi. https://t.co/we39LXMHiK
— Ethan Mollick (@emollick) August 12, 2024
Inspired by flight data recorders, it aims to identify and reduce preventable surgical errors, which cause up to 251,000 patient deaths annually in the U.S.
The first step is data collection.
In my experience teaching experts to use AI for work, a big change happens when they stop feeling relieved or smug when AI can’t do something and instead try to figure out how to make the AI succeed
Ask the question: “if this was a human, what feedback or guidance would I give?”
— Ethan Mollick (@emollick) August 11, 2024
Grantcharov installed the black boxes in about 20 U.S. operating rooms.
The Human Feeling of Artificial Intelligence. As described by the artists who are using it. pic.twitter.com/lZzQraMydb
— Runway (@runwayml) August 12, 2024
They capture up to 1 million data points daily per site, including audio-visual recordings, electronic health records, and the surgical team’s biometric readings. Next, AI excels at finding patterns in this data that the human brain cannot constantly monitor.
Grantcharov’s team hypothesized that surgeon stress affects cognitive processing and decision-making. By analyzing the data, they found stressed surgeons made more errors. They also identified distractions like doors opening or unrelated conversations as causes of catastrophic mistakes.
"In “The A.I. Mirror: How to Reclaim Our Humanity in an Age of Machine Thinking,” @ShannonVallor argues that we vastly underestimate our own richness compared with that of A.I. “ #books #AIEthics https://t.co/SvPgxP2fQC
— Carissa Véliz (@CarissaVeliz) August 11, 2024
Poor team communication and lack of psychological safety led to worse outcomes regardless of the surgeon’s skill. Once AI highlighted the biggest error sources, hospitals could implement procedures to prevent mistakes.
Reducing errors through AI analytics
However, changing the organization’s culture is key. Priorities like improving patient outcomes must be established. Desired behaviors need to become automated habits through training.
Systems should make following them easy. AI transformation also requires a growth mindset – viewing failures as improvement opportunities. Surgical teams were initially wary of the black box but shifted to recognizing the importance of objective performance measurements.
This led to a culture of continuous improvement. Though healthcare has high stakes, AI’s potential extends to other sectors. In meetings, it could analyze anonymized data to identify unspoken dynamics like people being talked over or showing stress.
Leaders could then foster more inclusive and productive environments. AI can unlock potential across organizations, but as technology plays a bigger role, understanding human interaction becomes more critical. Combined with behavioral insights, AI can drive meaningful change in operating rooms, boardrooms, and beyond.
