At the upcoming SAE World Congress, ATX will present results from a study they conducted with the Virginia Tech Transportation Institute (VTTI) showing that their own voice-based texting system reduces driver distraction when compared to manually texting. That should come as a shock to absolutely no one. However, VTTI has previously shown that manually texting while driving increased crash risk by 23 times over baseline driving. So based on these two studies, what can we conclude? Logically only that the crash risk of using a voice-based texting system is somewhere lower than 23 times that of normal driving.

That's simply not good enough.

There is ample research to suggest that voice interfaces reduce distraction compared to the comparable manual interface, and some studies, including some of my own primary research, show that for some measures driving performance using voice interfaces is indistinguishable from normal driving. But we still do not know the true effect on crash risk of using a voice interface, or how errors in voice recognition affect crash risk.

Naturalistic research, such as that carried out previously by VTTI (although I must note the ATX study carried out by VTTI was not a naturalistic study, but was instead conducted on their Smart Road), is the only available methodology for determining crash risk, which as I keep pointing out is the gold standard metric that all these researchers should be trying to determine. It's questionable whether voice interfaces have achieved the levels of usage among consumers that will allow us to determine that crash risk in the next naturalistic study.

So what can be done?

Well, first would be to design better studies. The ATX study is lacking in that it doesn't compare voice interfaces to baseline driving, or to any other task for which it has been determined that crash risk is raised but by less than 23 times. That can at least provide some rough guidance as to where the risk lies.

OnStar has been the only company to date that has had the courage to put its own internal data to the test. Of course for full disclosure I worked on that project and co-authored that study, but it was surprising how willing they were to take a big risk. Other companies, if they truly believe in and stand behind their product, should follow suit and support more rigorous research that will contribute to the discussion on driver distraction.

Chris Schreiner

Links:

ATX press release on study

OnStar study on crash risk

ACI research on driver distraction