Safety Expectations Too High for Self-Driving Cars?

Automated cars should help reduce crashes by being more alert than human drivers are. But researchers at the University of Michigan Transportation Research Institute say the safety expectations for self-driving cars may be too optimistic.

Analysts Michael Sivak and Brandon Schoettle point to several areas of concern. For one, they note that some crashes are due to vehicle failure, not human error. They opine that the complexities of technology for self-driving cars could make such vehicles more likely than conventional vehicles to experience an equipment failure.

They also observe that human drivers acquire predictive skills as they gain experience on the road. The researchers question whether such capabilities can be programmed into a computer-controlled car sufficiently to surpass the skill of today's safest group: middle-aged drivers.

Finally, the UMTRI researchers fret over the dynamics of the decades-long transition period during which conventional and autonomous vehicles will share the road. They note that drivers often rely upon eye contact with each other to determine how to proceed. But when self-driving cars and mixed together, they theorize, the lack of such feedback could make driving conventional vehicles less safe.