Beth Israel Labor Nurses Welcomed an Unusual Co-Worker
A robot developed by MIT helped nurses make tough decisions.
Nurses in the labor ward at Beth Israel Deaconess Medical Center welcomed an unusual co-worker recently: a robot developed by MIT.
Researchers from MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) created algorithms for a robot named Nao that allow it to learn by observing human decision-making, and supports workers by helping them make tough choices. CSAIL tested Nao at Beth Israel, using the robot to help nurses with mentally taxing scheduling jobs, who are showered with an endless stream of complex decisions—sometimes coordinating 10 nurses, 20 patients, and 20 rooms at the same time.
Julie Shah, an MIT professor who was the lead author on a research paper about the robot, says there are a few naturally gifted scheduling experts in the field, but it’s hard to train others. Nao is able to help those who struggle by offering potential good and bad decisions.
“The robot is never replacing the person doing the job, only supporting,” Shah says. “Similar to an apprenticeship.”
Nao learns how to perform the scheduling job similarly to how a human would: through observation. It starts with a lab simulation, getting a snapshot of the entire context—where rooms are located, where nurses can be assigned, and which patients need to be served. When in the hospital on a later day, Nao observed the labor ward, then got to work.
In testing, nurses accepted Nao’s recommendations 90 percent of the time. Nao was also shown to distinguish between good and bad suggestions when asked by nurses.
According to a CSAIL statement, nurses had almost unanimous positive feedback. One explained that Nao would allow for a more even workload among the nurses, so those gifted in scheduling aren’t carrying all of the weight.
“This was a mic drop finding, a success in the research world,” Beth Israel obstetrician Neel Shah says. “Usually when you have human experts, they have no trust in robots. However, nurses built trust because it was so good at making recommendations.”
In a second paper, the CSAIL researchers applied the same system to a video game that simulates missile-defense scenarios. The system occasionally outperformed human experts at reducing both missile attacks and the cost of using decoys to ward off enemy attacks.
“These initial results show there is tremendous potential for machines to collaborate with us in rich ways,” Julie Shah said in a statement.