Minding Your Busyness

MOST PEOPLE AREN’’T AS BUSY as they think they are, which is why we can usually tolerate interruptions from our inconsiderate electronic paraphernalia. James Fogarty and Scott E. Hudson of Carnegie Mellon University recently teamed up with Jennifer Lai of IBM Research to study 10 managers, researchers and interns at work. They videotaped the subjects and periodically had them rate their “interruptibility.” The amount of time the workers spent in leave-me-alone mode varied from person to person and day to day, ranging from 10 to 51 percent. On average, the subjects wanted to work without interruption about one third of the time. In studies of Microsoft employees, Horvitz has similarly found that they typically spend

more than 65 percent of their day in a state of low attention.

Today’s phones and computers, which naively assume that the user is never too busy to take a call, read an e-mail, or click “OK” on an alert box, thus are probably correct about two thirds of time. (Hudson and Horvitz acknowledge, however, that it is not yet clear how well these figures generalize to other

jobs.) To be useful, then, considerate systems will have to be more than 65 percent accurate in sensing when their users are near their cognitive limits.

Fortunately, this doesn’t seem to require strapping someone into a heart monitor or a brain scanner. Fogarty and his collaborators have found that simply using a microphone to detect whether anyone is talking within earshot would

raise accuracy to 76 percent. That is as good as the human judgment of coworkers who viewed videotapes of the subjects and guessed when they were uninterruptible. When Fogarty’s group enhanced the software to detect not only

conversations but also mouse movement, keyboard activity and the applications

running on machines, the system’s accuracy climbed to 87 percent for the two managers. Curiously, it rose only to 77 percent for the five scientists, perhaps because they are a chattier bunch.

Bestcom/Enhanced Telephony, a Microsoft prototype based on Horvitz’s work, digs a little deeper into each user’s computer to find clues about what they are up to. Microsoft launched an internal beta test of the system in mid-2003.

By last October, Horvitz says, about 3,800 people were using the system to field their incoming phone calls.

Horvitz himself is one of those testers, and while we talk in his office in Redmond, Wash., Bestcom silently handles one call after another. First it checks whether the caller is listed in his address book, the company directory, or its log

of people he has called recently. Triangulating these sources, it tries to deduce their relationship. Family members, supervisors and people he called earlier today ring through. Others see a message on their computer that he is in a

meeting and won’t be available until 3 P.M. The system scans Horvitz’s and the caller’s calendar and offers to reschedule the call at a time that is open for both. Some callers choose that option; others leave voice mail. E-mail messages get a

similar screening. When Horvitz is out of the office, Bestcom automatically offers to forward selected callers to his cell phone—unless his calendar and other evidence suggest that he is in a meeting.

Most large companies already use computerized phone systems and standard calendar and contact management software, so tapping into those “sensors” should be straightforward. Not all employees will like the idea of having a microphone on all the time in their office, however, nor will everyone want to

expose their datebook to some program they do not ultimately control. Moreover, some managers might he tempted to equate a “state of low attention” with “goofing off” and punish those who seem insufficiently busy.

The researchers seem to appreciate these risks. Hudson argues that an attentive system should nor record audio, keystrokes or the like but simply analyze the data streams and discard them after logging “conversation in progress,”“typing detected,” and so on. “We built a privacy tool into Bestcom from the beginning,” Horvitz emphasizes, “so users can control who is allowed to see the various kinds of information it collectsabout them.”

Watching the Watcher

AS DIGITAL CAMERAS fall in price, that information may come to include video. With a simple $20 webcam, Horvitz’s software can tell when a person is in view and whether she is alone or in a meeting. Fancier cameras can use the eyes as a window to the mind and perhaps extend the reach of considerate computers into the home.

Vertegaal has filled the Human Media Lab at Queen’s University with everyday appliances that know when you are looking at them. “When I say ‘on,’ the lamp over there doesn’t do anything,” Vertegaal says, pointing over his shoulder. He turns to face the object.

“On,” he says. LEDs mounted on a small circuit board stuck to the lamp Shoot invisible infrared light into his pupils. The light reflects off his retinas, and an infrared camera on the board picks up two bright spots in the image, one

from each eye. A processor does some quick pattern and speech recognition, and the lamp switches on.

Gaze detection can endow quotidian machines with seemingly magical behavior. Vertegaal answers a ringing telephone by looking at it and saying “Hello.” When he stops talking and turns away from the phone, it hangs up. The TV in the lab pauses a DVD or mutes the sound on a broadcast show whenever it notices that there are no longer any eyes watching it. Some of Vertegaal’s students walk around with eye-contact sensors on their hat or glasses. When the wearer enters a conversation, the sensor passes that Information via a wireless link to the cell phone in his pocket, which then

switches from ring mode to vibrate.

Although the technology is steadily improving, gaze detectors are still too expensive, bulky, ugly and unreliable for everyday use. “Eye contact is the most accurate measure of attention that we have—about 80 percent accurate in conversational settings,” Vertegaal says. “But it’s not perfect by any means.”

Attentive appliances are mere parlor tricks, moreover, when they act independently. The real payoff will only come from larger, smarter systems that can both divine the focus of our attention and moderate our conversation with all

our personal machines. Doing that reliably will require a nice bit of reasoning.

Date: 2015-01-29; view: 754