As I discovered the first article, I forgot the original article from Tom Simonite which is more publicly available. Again, some interesting thoughts bubble out of this article, which I can agree with his findings – especially when being the pilot of the robotic telepresence.
Our elastic sense of personal space posed one of the biggest problems. At a crowded party people get more cosy, and that seems to apply to robots too. Unfortunately the robot is only around four feet tall, and has a camera with a relatively narrow field of view. Even though I could tell that I was in a group conversation, I could only see one person at a time, from the viewpoint of a child, a strange experience for someone who stands 6 feet tall. Threading my way around the crowed party without bumping into things was a challenge I didn’t master.
One of my personal concerns as I mentioned before – height is very important since our sense of intimacy and position is often determined by the stature of the participants in a conversation. And having a wider field of view (humans are more than 180 degrees, webcams are often 40 to 60) is extremely important. But it is important to note that while we have great peripheral vision, we tend to focus on one area at a time – while the rest of the field of view is fuzzed out and rapidly focus as we move from one viewpoint to another (try this yourself).
Talking was made difficult because when you’re a robot you miss out on human brain’s powerful Cocktail Party Effect that can pick out a single voice from the hubbub of many others. The robot’s multiple microphones delivered voices from all angles into my ears at high quality, something that’s valuable in the boardroom, but confusing at a party. The combination of background chatter and the people I wanted to hear talking over one another or to people passing by made for a dense wall of sound.
We found this a major problem as well when we had John Markoff using our first version of the Texai at the PR2 Launch party. Low light is also a problem (as John learned) and the audio problem is something that needs to be addressed.
The best social interaction I had at the party was with a child of around five years old. Not only was she much closer to “my” eye level, but unlike a grownup she was also happy to devote her entire attention to our conversation and ignore everyone else. She also seemed to find it less of a novelty to be talking to a robot—indeed she wasn’t interested in talking about my odd appearance at all.
And as we experienced, kids loved using the Texai – both as a pilot and as a participant. I had one of my friend’s son come into the office and sat him in front of the computer with the old interface exposed. Next thing I knew, the Texai was tooling around the office at a fast clip. When his father saw him and spoke with him, it was as natural as standing in front of him.
Neat post and thoughts. I wonder if a dual-camera setup could be worth trying, with one camera having a wide-angle lens to get a peripheral vision simulacrum…
The height issue seems important and solvable. The audio issues sounds like a bigger challenge…
Actually, the RP7i from InTouch Health is offering a solution with dual cameras – see it at http://www.intouchhealth.com/products_rp-7_robots.html.
The height issue is solvable – though it has associated cost and safety issues. You can see how iRobot handled it on their AVA.
As for the audio, everyone suggests some form of microphone array – but there is another set of challenges there that require a great deal of thought.