You'll never understand your robot until you walk a mile in his screws. A new MIT study reveals that when co-working humans and robots cross-train, productivity improves.

Robots already perform rote tasks and heavy lifting alongside humans in the manufacturing world. But increasingly, robots are partnering with people in places as complex and critical as the operating room and outer space, says Stefanos Nikolaidis, a Ph.D. student who co-authored the study with Julie Shah, head of MIT's Interactive Robotics Group.

In the study, a human and a robot (a big orange mechanical arm) had to work together on a task. The human placed screws in holes, and the robot drilled them in. Then, the pair swapped roles, via a computer simulation. Afterward, the pair returned to their original real-world tasks.

The scientists designed an algorithm that let robots perceive human preferences (for example, whether screws should be drilled immediately or whether they should be done in batches).

The cross-trained teams were 71 percent more productive than teams that used other training strategies. Cross-trained humans also reported a higher level of trust in their wired partner.

"Whether we are talking about human teammates or human-robot teams," says Nikolaidis, "jobs are done better when there is understanding and trust between teammates."