Robots Learn To Look Shifty
Writing by Evan Ackerman on Tuesday, 3 of November , 2009 at 12:20 am
Robots are slowly learning to take advantage of nonverbal cues, which are an integral part of natural human communication. Things like eyebrow movement can be used to convey information about emotional states, but subtle nonverbal cues can also provide more pointed information… Think about how much can be said with a quick glance. Professor Bilge Mutlu from the University of Wisconsin is trying to teach robots to communicate information with glances, by seeing how well effectively they can “leak” the answer to an object guessing game played with a human.
As you can see from the video, two robots took place in this experiment: Robovie R-2, and the creepy as f**k Geminoid. Interestingly, participants in the experiment responded better to cues leaked by Geminoid, while at the same time, the didn’t realize that they were doing so, implying that however deep in the Uncanny Valley Geminoid may reside, in some cases, having a robot with distinctly humanoid features really does make a difference.
There is one itsy bitsy teeny tiny little caveat at the end of the paper: “We found that the leakage cue affected the performance of only pet owners and not others, which might suggest that pet owners become, through their interaction with their pets, more sensitive to nonverbal behavior.” Er, okay, well, I guess it was good for them to check for that kind of thing, but that kinda makes the effects of shifty-eyed robots a lot less impressive.
[ Paper (*.PDF) ]
Leave a comment
Category: Research, Uncanny Valley
- Add this post to
- Del.icio.us -
- Digg
No comments yet.