If your job is in manufacturing, medicine, mining, automotive repair, underwater or space exploration, maybe even elder care, some of your coworkers are probably semi-autonomous programmable mechanical machines--in a word, robots. But humans and robots don't understand each other well, and they work very differently: a robot does exactly what it's told, over and over, taking the same amount of time every time, while a human acts with intention, deliberation, and variation. Their strengths and weaknesses can be complementary, but only if they each have good models of how the other works.
In a breakthrough experiment, the Interactive Robotics Group at MIT discovered that cross-training, which is swapping jobs with someone else on your team to help everyone understand the work better, works even when your coworker doesn't have a mind. In short, when humans and robots model doing each others' job they end up working together more smoothly.
In order for this to work, researchers first had to program robots to learn by watching humans instead of just through feedback. Humans were paired with a robotic arm, named Abbie, to practice placing the screws and screwing them in--in a virtual environment. There were two basic rhythms to the task: either have Abbie fasten the screw right after it was placed (1/2, 1/2, 1/2), or place all three screws and then have Abbie screw in the batch of three (1-2-3, 1-2-3). After the humans modeled their actions, and the robots practiced placing the screws, the team moved to a real environment where humans placed screws and Abbie screwed them in.
The outcome was fascinating. In the control group, the humans and robots move like awkward dance partners. The human isn't sure where the robot will go next, and doesn't want a screw driven through her hand, so she spends more time waiting around while Abbie is moving.
The team that had cross-trained understood each others' preferences much better. They spent 71 percent more time moving at the same moment, a sign of better coordination--like a well-oiled machine, you might say. The humans spent 41 percent less time waiting around for the robot's next action. The humans rated the robot's performance higher, and the robots had a lower "entropy level," meaning they spent less time in uncertainty about what the humans would do next.
"What we suspect, and are planning to follow up on, is that the real benefit is coming from adaptation on the human side," said MIT professor Julie Shah, who leads the Interactive Robotics Group. "The person is doing actions in a more repeatable way, developing a better understanding of what the robot can do."
Cross-training is a technique the military uses to improve teamwork. How could it work for you?
[Robot and Human Image: Daniel Schweinert via Shutterstock]