I am not a natural team player: I hate to have to rely on other people performing well but, equally, I am devastated if I fear I have let them down. That makes working with me a tough brief for my latest teammate, Abbie, a bright orange, industrial robot arm. Together we are going to insert three screws into a tabletop.
It seems simple but, behind-the-scenes, a profound human-machine connection could emerge. I just need to meld my mind to her software brain.
Whether humans and robots can bond as teammates used to be a theoretical concern
. Factory robots tend to sit in cages, safely cordoned off from their human colleagues, so the two cannot touch, let alone collaborate. That is starting to change.
Baxter, a robot that can work safely beside humans, was launched by Rethink Robotics of Cambridge, Massachusetts, at the end of last year. Meanwhile, the US National Institute of Standards and Technology is creating safety standards for us to share workspace with robots by 2014.
To fully exploit this opportunity, Stefanos Nikolaidis, one of Abbie's programmers, and his supervisor, Julie Shah, want to teach robots to work fluidly with people and be better at anticipating their actions – as well as winning their trust. To do this, they have turned to human psychology.
My collaboration with Abbie will be based on a psychological technique called cross-training, in which members of a team work more effectively if they swap roles in preparation for a task. The rationale is that they will converge on a shared plan – or mental model – to complete the task, leading to fewer errors on the day.
Will the same be true for Abbie and me? From the beginning, the chips are stacked against us – deliberately.
There are two ways to perform our task: Abbie can either get to work as soon as I have put the first screw in the hole, or wait until I have them all in place and then screw down all three in one go. I express a preference for the first option, so Nikolaidis loads Abbie's software brain with the exact opposite preference – to ensure the mind-meld is as hard as possible to achieve.
We rehearse in a virtual workshop and swap roles so I control a virtual Abbie wielding a screwdriver and she controls a virtual me. When she puts a screw in place I jump in to screw it down before she has moved onto the next one, even though she's programmed to expect me to wait.
When it is time to try the task for real, Nikolaidis uploads the updated version of Abbie's brain into the robotic arm and I don a black glove covered in red LEDs to make it easier for the robot to read my hand movements. We stand facing each other across a bench top drilled with three holes.
I know I have adapted as a result of the rehearsal but has Abbie? Abbie's machine-learning software analysed my behaviour as I carried out her role, which gave her a glimpse of my expectations of her. Because I performed it differently to the way she had been programmed, she should have modified her behaviour to meet my expectations. "The goal becomes not only to do the task, but in a way that is closer to the person's preferences," says Nikolaidis. We start the task Abbie's way but end up completing it in a hybrid fashion – she starts to do things my way once I've positioned the second screw.
I am not the only human capable of bonding with a robot. Shah and Nikolaidis asked 18 volunteers to cross-train with virtual Abbie too. Those who cross-trained also stood closer to Abbie, spent more time moving at the same time as her – a sign of fluidity that perhaps arose from greater trust – and said they believed that Abbie had "learned their preferences". The results will be presented on 4 March at the Human-Robot Interaction conference in Tokyo, Japan.
Disappointingly, I didn't feel an eerie connection to my robo-teammate, but perhaps that is a sign of how well the training worked. "The fact that you weren't thinking about it at that level – or stressed out – that's a success of cross-training too," says Shah.
If you would like to reuse any content from New Scientist, either in print or online, please contact the syndication department first for permission. New Scientist does not own rights to photos, but there are a variety of licensing options available for use of articles and graphics we own the copyright to.
All comments should respect the New Scientist House Rules. If you think a particular comment breaks these rules then please use the "Report" link in that comment to report it to us.
If you are having a technical problem posting a comment, please contact technical support.
Trading places with us makes robots better teammates
This article
Trading places with us makes robots better teammates
can be opened in url
http://newseristical.blogspot.com/2013/02/trading-places-with-us-makes-robots.html
Trading places with us makes robots better teammates
