A collaboration between Toyota and Stanford
University is experimenting a talking car interface that does much more than
navigation, providing safety advice to drivers. Ing-Marie Jonsson is one of the
authors of the paper – presented today at CHI 2008 – that describes the
experiment. She gives us more information about their work in the following
interview:

What can a "talking car" do for its driver, besides the usual
navigator functions?
”The future of in-vehicle information systems will include personalized access
to different driving and convenience related information sources, weather,
traffic, news food, shopping, activities, special events etc. and
entertainment/work such as music , videos, and internet activities (web,
email). Social networking functions will enable people to meet and socialize
based on common interests. More uniquely, the GPS ability of the car
furthermore enables social networking based on common locations. “

Could a talking car cause “undesired side effects" in driver’s behaviour?
”Yes. While it’s different to a passenger, if a talking car is not designed
properly it can distract the driver. Talking when the driver is in a difficult
driving situation, or interrupting the driver, or simply providing incorrect
information (thus making the driver frustrated) – can all cause the driver to
exhibit decreased driving performance and a negative attitude towards the
talking car.”

How did you study users to understand the effects of your talking prototype on
them?
”We design experiments using human participants that interact with speech based
in-vehicle information systems while driving a car simulator. This gives us
precise control over both the speech system and the driving environment. To
study situations when drivers show bad judgment and inattention due to the
speech systems, we design driving scenarios that are often hard and full of
traffic situations.”

Did you find any surprising behaviours?
”Dealing with an in-car speech system is different than dealing with one in a
lab or office, and is most definitely different than dealing with other human
passengers.  We have found differences in behaviours ranging from gender differences to age
differences. For example, certain high levels of inaccurate information can
cause male drivers to drive extremely badly, whereas female drivers remain
largely unaffected. Both genders correctly determine how ccurate the system is,
however female drivers are appear to be far less affected by this. Not
surprisingly however, we found both gender stereotyping (navigation
recommendations more believable using a male voice) and higher trust and liking
of same-gender voices.
Another interesting find is that the car is a special environment for older
drivers. In a study investigating how the age of voice impacted driving
behaviour of older adult and young drivers, we found that all drivers prefer
the young voice in the car. Older adult drivers like and trust the young voice
more than the older adult voice when it provides hazard and warning information
in a car. This is surprising since when the same information is given in an
office setting; older adults prefer and trust the older adult voice over the
young voice.
We have also seen how a system that is too attractive and fun can be
disastrous, a recent study showed that when a conversational system in the car
was too interesting to communicate with, drivers forgot about driving with
disastrous consequences.”

How do you imagine in-car information systems of the future?
”Including all the features that I mentioned before, we see a well designed
in-vehicle information systems a system that is helpful, without being
intrusive, a system that is fun and engaging without moving the drivers’
attention away from the driving task. This requires attention to design and a trade-off in attractiveness vs being
ignored.”

© 2008, Il Sole 24 Ore. Web report from CHI 2008.