My last area of endeavour at AT&T was on the Augmented Vehicles project, where we were designing more intelligent in-car systems to try and make time spent on the move more productive, safe, educational and enjoyable.

One of the ideas we explored was context-sensitive phone management...

The reason, we suggested, that it's measurably more dangerous to talk on the phone while driving than it is to talk to a passenger sitting next to you, is that the person at the other end of a call doesn't share the same visual cues as the driver.

If you're driving and I, as a passenger, ask you a question just as you are negotiating a difficult junction, I'll understand completely if you don't reply immediately. If I'm on the other end of a phone line, however, you'll feel more compelled to maintain a normal conversation with me and so will devote more attention to that than perhaps you should.

Our idea was that we could make use of the large GPS logs we had gathered over the years to improve this situation. If I called you just as you were approaching a potentially difficult point - a complex junction, a tight bend, somewhere which typically had dramatic speed changes or tight radii of curvature - I could be informed that you were driving and could answer in 20 seconds. The phone wouldn't even ring in the car until you were onto easy, straight roads.

That at least was the idea. Sadly, this project was short-lived, starting only a few months before the rather unexpected demise of the Lab.