New Technology Focus: Computers that know you're upset!

Imagine this scene: You're sitting in front of your computer, working away, when it does something really annoying -- a scheduled maintenance program starts itself while you're in the middle of writing that proposal, your browser stops to clean the cache when you're desperately looking for a file on the Internet, the auto-correct changes your boss's name from Webb to Web.

Now imagine that after this happens a few times, your computer figures out that it's annoying you, and stops doing whatever was driving you nuts.

Sounds like science fiction, right? But researchers at MIT are studying ways for the computer to do just that, and more.

The Affective Computing Department at the MIT Media Lab in Cambridge, Mass., is trying to use computers to detect people's emotions. With that information in hand, the computer can change its behaviour to suit you, or pass that information on to others. "The computer could actually adapt to your behaviour. If it does something that drives you nuts it doesn't always have to do that," said Rosalind Picard, an associate professor of media technology at MIT. Computers are meant to be our assistants, she said, and they should learn our likes and dislikes just as a human assistant would.

"Computers act more and more not just as tools, but as agents that you delegate tasks to and ask to perform tasks based on your preferences," she said. There are a few steps to get through before computers are ready to cater to our emotional needs, though. First, they have to be able to figure out what you're feeling. Most of those tests involve wiring up volunteers to monitors that check for things like heart rate, blood pressure and skin conductivity, then putting them through a series of tests. Not surprisingly, the emotions the researchers are looking for are usually anger and frustration.

One test involved asking volunteers to click on the correct answer as fast as they could, with the winner getting a $100 prize. What they didn't know was that the mice had been rigged to "stick" and slow down. Another testing device involves attaching a sensor pad to a mouse button that can detect extremely small muscle twitches. People have tendencies to pull things they like toward them, and push away things they don't -- a tendency that can be monitored when you're clicking through a Web site, for example. That type of information could be useful to a designer testing a new Web site.

Another product that could help designers is a set of glasses that can monitor movements in the eyebrow muscles. Basically, the glasses detect whether you're furrowing your brow out of confusion. That data is sent out to a computer, which displays a bar chart monitoring the level of confusion. Volunteers could wear them while testing a new product, so the designer could tell exactly when something wasn't working or was puzzling to the tester.

The glasses could also be used for teaching. One example: Remote education, in which the professor is in one place and the students are somewhere else. "You could just point a camera at someone's face and use pattern recognition to determine confusion. But people in classrooms don't always show confusion because they don't want to be perceived as less intelligent," Picard said.

If the entire class were wearing the glasses, the professor would know in an instant when something wasn't getting through.