When Mangesh Gururaj's wife left home to pick up their child from math lessons one Sunday earlier this month, she turned on her Tesla Model S and hit "Summon," a self-parking feature that the electric automaker has promoted as a central step toward driverless cars. But as the family's $65,000 sedan reversed itself out of the garage, Gururaj said, the car abruptly struck the garage's side wall, ripping its front end off with a loud crack. The maimed Tesla looked as if it would have kept driving, Gururaj said, if his wife hadn't hit the brakes. No one was hurt, but Gururaj was rattled: The car had failed disastrously, during the simplest of maneuvers, using one of the most basic features from the self-driving technology he and his family had trusted many times at higher speeds. "This is just a crash in the garage. But what if we were summoning and there was a child it didn't see?" said Gururaj, an IT consultant in North Carolina, who bought the car last year.

Network and other computer administrators typically have access to a rich set of logs tracking actions by users. However, they often lack metadata such as user role, age, and gender that can provide valuable context for users' actions. Inferring user attributes automatically has wide ranging implications; among others, for customization (anticipating user needs and priorities), for managing resources (anticipating demand) and for security (interpreting anomalous behavior).