A Phone that Knows You're Busy It's a modem conundrum: you're too busy to be disturbed by incessant phone calls so you mm your cell phone off. But if you don't remember to turn it back on when you're less busy, you could miss some important calls. If only the phone knew when it was wise to interrupt you, you wouldn't have to turn it off at all. Instead, it could let calls through during spells of relative inactivity. A bunch of behavior. sensors and a clever piece of software could do just that, by analyzing your behavior. to determine if it's a good time to interrupt you. If built into a phone, the system may decide you're too busy and ask the caller to leave a message or ring back later. In a desktop computer, the system could stop instant messages or spain annoying you when you're busy. James Fogarty and Scott Hudson at Camegie Mellon University in Pennsylvania based their system on tiny microphones, cameras and touch sensors that reveal body language and activity. First they had to study different behaviors to find out which ones strongly predict whether your mind is interrupted. The potential 'busyness' signals they focused on included whether the office doors were left open or closed, the time of day, if other people were with the person in question, how close they were to each other, and whether or not the computer was in use. The sensors monitored these and many other factors while four subjects were at work. At random intervals, the subjects rated how in term ptible they were on a scale ranging from 'highly interruptible' to 'highly not—term ptible'. Their ratings were then correlated with the various behaviors. 'It is a shotgun approach: we used all the indicators we could think of and then let statistics ferret out which were important,' says Hudson. The model showed that using the keyboard, and talking on a landline or to someone else in the office correlated most strongly with how interruptible the subjects judged themselves to be. Interestingly, the computer was actually better than people at predicting when someone was too busy to be interrupted. The computer got it right 82 per cent of the time, humans 77 percent. Fogarty speculates that this might be because people doing the interrupting are inevitably biased towards delivering their message, whereas computers don't care. The first application for Hudson and Fogarty's system is likely to be in an instant messaging system, followed by office phones and cell phones. 'There is no technological roadblock to it being deployed in a couple of years,' says Hudson. What is the modem conundrum the author has in mind?