With the help of an advanced robot, scheduling at hospitals across the country may soon be easier than ever before.
The robot, created by a team at the Massachusetts Institute of Technology’s Computer Science and Artificial Intelligence Lab (CSAIL), will be designed to help schedule and assign tasks in a wide range of settings, including both medical and military. With artificial intelligence, the robot gathers knowledge by “watching” experts perform tasks in the field and then recommending similar solutions based on what it has seen.
Once the robot is ready to perform, it can basically do the job of an “air traffic controller” for these workplaces, said Julie Shah, associate professor in MIT’s Department of Aeronautics and Astronautics and senior author on two new papers describing the team’s research.
“There are very, very few people who do these hard jobs very well,” Shah said. “The idea was, ‘Can we learn from these truly exceptional experts at the task?’ and if so … it can help us make a training tool to help novices,” she added.
The robot was put to work professionally for the first time at Beth Israel Deaconess Medical Center’s Labor and Delivery ward. The findings from that trial were presented at the International Joint Conference on Artificial Intelligence in mid-July.
“Figuring out what makes certain people good at [scheduling] often seems like a mystery,” PhD student and co-author Matthew Gombolay said in a statement from MIT. “We thought a complex environment like a labor ward would be a good place to try to automate scheduling and take this significant burden off of workers.”
Robots in hospitals aren’t exactly a new idea, but up until MIT’s latest development, most robots have only been given menial tasks, such as delivering towels or medicine to patients.
But these robots are often unpopular, Shah said. Because they need “painstakingly” specific instructions, the robots often create work for doctors and nurses. CSAIL’s new robot, however, only requires a few dozen demonstrations before it can perform on its own.
“That’s the game-changer with this, being able to learn much more efficiently with fewer demonstrations,” Shah said.
The results of the trial run at Beth Israel, which lasted from November 2015 through January, were even better than expected, Shah said. Ninety percent of the advice given by the robot was accepted by doctors and nurses as “high quality advice.”
Additionally, the nurses at the hospital who interacted with the robot “had almost uniformly positive feedback about the robot,” MIT’s release stated.
The next step for CSAIL is to determine whether the robot can learn “generalizable strategies” that can be applied to a diverse set of hospitals and settings across the country. The team at MIT is collaborating with other researchers in the region, including a team at Harvard University.
“The larger goal behind the work we’re doing in the lab is to create human-aware AI [artificial intelligence],” Shah said. “We take on challenging research projects purposely, but for us, this is probably one of the most challenging research projects we could pursue.”
Photo Credit: Massachusetts Institute of Technology’s Computer Science and Artificial Intelligence Lab
Felicia Gans is an editorial intern at MedTech Boston. She will be a senior this fall at Boston University, where she is studying journalism, political science, and computer science. When she's not working, Felicia loves drinking coffee, jamming out to Broadway music, and reading the news.
Send this to a friend