Tip:
Highlight text to annotate it
X
Researchers at MIT's Computer Science and Artificial Intelligence Lab (CSAIL) have discovered that humans prefer increased autonomy for robots.
And no, those researchers were not robots.
More robot control, MIT’s research shows, leads to greater efficiency and helps people understand their robot counterparts. More autonomy, however, doesn’t mean full control; humans are still delegating and scheduling robots' tasks, but through a “human-generated algorithm.”
Researchers were looking for ways to integrate robot workers into factory areas traditionally filled with humans without making them “feel devalued or resistant to the technology," according to the video below.
"We need to understand how to gain the support of human workers who will use this technology," explains Matthew Gombolay, a Ph.D. student at MIT's lab, in a video detailing the work.
The goal of the research was to show that a balance between robot control and human supervision would result in greater team productivity and humans' acceptance of the robots.
That’s not exactly how it worked out.
In their experiment, MIT researchers brought together 24 humans to work alongside a two-armed Willow Garage PR2 robot in building Lego kits. Containers were filled with Lego parts that the robots could access, but they were not doing the building; the robots were adjacent to humans who could take the parts out of the bins to build with. But no two workers — robot or human — were permitted to be in the same area at the same time. (Workers couldn't go into the bin area when a robot was busy there.)
The team conducted three tests in which a robot had varying degrees of scheduling decision control: manual (human sets the whole schedule), semi-autonomous (some human, some robot) and fully autonomous.
It soon became clear that the more autonomy the robot received, the better the team performed.
There was, it seemed, no sweet spot between human control and robot autonomy. Both team performance and satisfaction of human workers is optimal when the robot is fully autonomous, the researchers found.
“We found that humans preferred working with the robot that had more control over scheduling decisions. Subjects felt that the autonomous robot better understood them,” Gombolay said.
These latest findings should be encouraging — at least, for robots. In a recent and widely shared YouTube video, CGP Grey lays out a case in which robots take over not only humans' repetitive, menial and dangerous jobs, but even ones in which deep thought and contemplation are required. If there is truth to the findings from the MIT researchers, humans be willing participants in the takeover of robot overlords.
Th researchers' algorithm works to create on-the-fly plans and even reschedules tasks when necessary. A human may have written the program, but now the robot — also built by humans — is eventually flying solo, finishing the task, making decisions and slowly but surely making us obsolete.