MIT World: Engineering >> 
While automakers market increasingly intelligent cars, they may be missing the point. No matter how sophisticated the vehicle’s brain, suggests
Alex (Sandy) Pentland, the smartest element on the road is still the human driver. In search of safe, responsive vehicles, designers should not think of separate components -- machine and operator -- but rather, an integrated system comprised of two, complementary intelligences.
Tackling this challenge involves analyzing human behavior -- not the traditional purview of engineers who “are scared off by the noise, the randomness of people.” Pentland, on the other hand, has long explored human decision-making in a variety of settings, including car driving. Working with an automaker, he outfitted test vehicles with sensors on the steering wheel, brakes and elsewhere, to “determine predictive signals of driving.” The sensors permitted the analysis of “clusters of behaviors,” so Pentland could figure out with 95% accuracy “what people would do before they did it.” From developing a model of the driver’s behavior, he moved on to neighboring drivers’ patterns, to paint a picture of road interactions and signaling.
This research has resulted in a system now