A version of this article was published in Risk & Insurance Magazine

We were in the basement garage of a big insurer. Every parking space was full. We raced the car down narrow lanes, braked sharply, swerved at tight corners. The driver of this manic jitterbug in my Mazda was a senior executive of the insurer, who a chance observer might surmise had lost his marbles. Luckily there were no pedestrians to see us or be run over. Nor did there appear to be surveillance cameras on the walls. The camera, as it happened, was in the car.

We were doing our own goofy pilot of a device that might drastically improve risky driving behavior, cut accidents by half, cut maintenance costs and bring down auto insurance premiums. If this kind of technology works, how can we get this device out of the basement, into commercial vehicles and every risky driver's car?

These accidents put 45,000 people into the grave each year. Safety researchers tend to focus on car design, impact-ignited protections and road design. But by detecting a few bad habits and providing feedback to the driver, it appears possible to improve overall behavior and lower the risk of accidents. Here is a practical solution to capturing these events. I had installed in my Mazda a device made by DriveCam, a San Diego-based firm. You stick it near your rear-view mirror. It records 10-second looping videos through front and rear windshields, with audio. When the driver crosses a variable threshold of braking and veering, the device retains these 10 seconds and the next 20 seconds. Liberty Mutual's top expert on driver safety, David Melton, likes it. Commercial fleets began installing DriveCam devices several years ago.

I visited DriveCam customer BostonCoach, one of the largest providers of ground transportation services worldwide. Director of Driver Training Tom Kenny allowed me to join him as he reviewed DriveCam footage. Kenny reviews the incidents weekly and sends comments to drivers. He may personally review the footage with the driver, recommend improvement or, if necessary, take disciplinary steps.

BostonCoach sets a low threshold to trigger the device. Most of its recorded incidents are caused by rough road conditions. In fact, very few (one out of eight) involve accidents or instances of deficient driving by his drivers or another vehicle.

In one video clip, a sedan was traveling down an empty country road that curved to the right. To my eye, everything seemed normal. Kenny, however, spotted a depression in the road, which he said triggered the device as the car went over it. He also pointed out that the car was traveling at a speed above BostonCoach's acceptable limit for that type of road. Kenny made a note to review the BostonCoach standard for operating in these types of conditions with the driver.

ACE, the insurer, found in a DriveCam pilot a 50 percent reduction in accident frequency, 79 percent reduction in accident severity, 22 percent reduction in cost-per-mile of operation, and 88 percent reduction in number of individual maintenance/repairs exceeding $4,000.

A DriveCam executive who used to run the Mayo Clinic's transportation fleet related an informal pilot of the technology with teenage drivers. They carefully avoided triggering the device. Some didn't want the reviewer (or their parents) to learn what music they listened to or with whom they drove! Incentives seem to work, whatever their color: BostonCoach drivers earn several hundred dollars a year by driving without an at-fault incident.

Didn't Hamlet say something about holding a mirror up to ... your driving? He was prescient. Let's take his advice.