Redefining
'Personal' Computing
Is teaching emotions
to computers the key to a sci-fi reality?
(Bright Magazine, march/april 2001)
courtesy of IBM |
| IBM's
Emotion Mouse can determine its user's emotional state by
tracking heart rate and skin temperature. |
In
this story:
A
Dog's Life
Big Brother Watching?
Related Links
By
Jonah Freedman
You've
had a rough day. Your meeting with the CEO was cancelled, your plane
was delayed three hours, and you can't get that horrifically catchy
*NSync song out of your head. So at the end of the day, who to turn
to?
Why, your understanding and empathetic computer, of course.
As preposterous as it might sound, if researchers across the country
are on the mark, the machines of the 21st century will be a lot
more understanding of our needs.
Possibilities include car-key remotes that know you're afraid while
walking through an empty parking lot at night and can contact authorities;
steering wheels that can sense when truckers are getting drowsy
and can recommend the driver pull over; or robotic pets that know
you need a little extra attention after a bad day.
"Affective computing" is the school of study aimed at developing
ways for computers to interpret our emotional states through basic
human cues, such as facial expressions, tone of voice and changes
in the autonomic nervous system. Researchers at IBM, Eastman Kodak,
the Massachusetts Institute of Technology, Stanford University and
the University of California, San Diego are just a few players in
this revolutionary field.
"If we want computers to be genuinely intelligent, to adapt to
us and to interact naturally with us, then they will need the ability
to recognize and express emotions, to have emotions and to have
what has come to be called 'emotional intelligence,'" writes Dr.
Rosalind Picard in her 1997 book, Affective Computing.
Picard is the director of the Affective
Computing Research Lab at MIT, one of the leading research centers
in the field. At the heart of this research is the assertion that
in order for machines to truly serve us better, they must understand
us better.
This goes far beyond the traditional ones and zeroes of computer-ese,
instead taking into account human traits that we have been otherwise
unable to make computers understand. Picture Data, for example,
the android of "Star Trek: The Next Generation" fame who was unable
to comprehend human emotion.
A
Dog's Life
Behind the large, blue doors of room 217B of the UC San Diego Cognitive
Science Building, Professor Javier Movellan and his staff at the
Machine Perception Laboratory
are working on refining the ways in which machines recognize and
interpret human audio and visual cues.
Inside the lab, finding a human test subject with his head in a
vise is not an unusual sight. A camera closely focuses on his facial
movements, which translate to changes in pixel color on a computer.
Movellan and his staff have programmed that same computer to correlate
these visual cues with aural data based on the tone of the words
the subject happens to be speaking.
Movellan says what makes this research unique is his team ignores
traditional technology such as voice-recognition software—something
programmers have been trying to perfect for almost 40 years.
"From the '60s up until now, speech-recognition software has focused
only on signal processing," Movellan explains. "It gets rid of how
you say things, effectively eliminating the emotional aspects."
Instead, the MPLab is starting from scratch, paying attention to
more primordial interactions between living things.
"Take dogs for example," Movellan says. "They don't understand
words very well—maybe 10 to 12 at most—but they are excellent
at understanding how the words are spoken. They know you're angry,
so they'd better be careful. Or they know you're happy with them,
and maybe they'll get a treat."
In fact, much of the MPLab's current research aims to add this
type of perception capacity to robot pets, similar to Sony's robotic
dog, AIBO. Movellan
envisions putting smart cameras inside these 'droid companions.
"We're working on giving these machines the ability to recognize
you and your expression," Movellan says excitedly. "So the dog will
be able to say to itself, 'He doesn't look as happy as he did yesterday.'
And the dog can change its behavior accordingly."
Big
Brother Watching?
Meanwhile, at IBM's Almaden
Research Center in San Jose, Calif., researchers are turning
to more basic forms of user input—physiological ones. Researcher
Wendy Ark designed one of the facility's creations, the Emotion
Mouse.
This gadget, which is embedded with sensors, is designed to determine
users' emotional states by tracking the physiological reactions—such
as changes in skin temperature or heart rate—that occur when
human emotions fluctuate. Ark says the most effective use of this
technology so far has been in distance learning, where lecturers
reach students in a different geographic location over satellite
or Internet.
"The problem is that professors can't see the class, so the lack
of feedback is detrimental," Ark says. With the Emotion Mouse, a
computer will be able to tell an instructor if his students are
confused or anxious.
Both Ark and Movellan say there are larger implications over the
next century, including the ability to tell if someone is trying
to deceive us over e-mail and machines that can baby-sit children.
However, a more frightening implication could be invasion of privacy
so commonplace that you're not even safe from your local cash-machine
cameras.
"Let's say it recognizes your face and notices you look depressed,"
Movellan explains. "It could contact Amazon.com and let the server
know. So the next time you visit Amazon, you'll get a home page
that says, here's the best book to cure depression.'"
The drawbacks of this technology also bring to mind science-fiction
fantasies. At what point does the technology get too intrusive?
On one hand, Data's pointed questions about human nature are invitingly
amusing. On the other, machines are making our decisions for us,
a la HAL 9000 of "2001: A Space Odyssey" and Big Brother of Orwell's
1984.
"We have to think very carefully about how this technology is used,"
Movellan warns. "It can be very invasive if it's used improperly."
Related
Links:
|
