Paper Reading #19: Local danger warnings for drivers: the effect of modality and level of assistance on driver reaction

Comments
Evin Schuchardt

Luke Roberts

Reference Information
Title: Local danger warnings for drivers: the effect of modality and level of assistance on driver reaction
Authors: Yujia Cao, Angela Mahr, Sandro Castronovo, Mariët Theune, Christoph Stahl, Christian A. Müller
Presentation Venue: IUI 2010: Proceedings of the 15th international conference on Intelligent user interfaces; February 7-10, 2009; Hong Kong, China

Summary
This paper focuses on various local danger warning presentations while driving as it pertains to situation awareness (SA) theory. SA is broken up into three phases for the driver:
1. Perceiving elements in the environment
2. Comprehending what these elements mean
3. Predicting the future states of the environment

The researchers focused on two major factors in their experiments to test different warning presentations: modality and level of assistance. Modality affects the driver’s perception through auditory or visual warnings, and level of assistance affects decision-making through suggestions such as ‘change lanes’ or ‘brake.’

In their experiment they used simulation software that they put inside a Mercedes-Benz. They projected the scene of a one-way highway with two lanes and no other vehicles on the windshield of the car. They displayed visual warnings on a screen to the right of the steering wheel. Auditory warnings were heard through a PC speaker. They included four types of obstacles (broken-down vehicle, fallen tree, rock and lost cargo) that were described to the driver in terms of type, location and distance ahead. They tested visual + blinking red bar, visual + a beep, visual + speech and speech-only warnings. After running the experiment as described, they ran it again, this time including auditory suggestions to measure level of assistance. They also ran a baseline experiment where the drivers were given no warnings.

The researchers found that in all cases, drivers reacted better with the warnings. They found that auditory suggestions helped drivers switch lanes and brake faster. The users were also reported to like the assistance and believed it would be more useful in real-life driving. The researchers point out that these results do not match with results from earlier studies in which users preferred less assistance.

Image taken from the paper

As far as modality, they found that the best warning came in the form of a visual and speech combination and speech-only warnings gave the worst results. The researchers decided that the duration of the speech message was too long and could not convey the location of the object to the driver in the same way as a visual warning could. Based on a survey given after the experiment, users preferred a beep over a blinking red bar, and the visual warning without sound received the lowest score on average.

Discussion
Overall this was a good paper. The researchers were thorough in the experiments and their explanations were easy to understand.

Something really interesting about this paper was how well it related to things we’ve been reading about in class. At the beginning of the paper, the researchers discussed how auditory warnings were thought to be better than visual ones because driving makes great use of perception. They also discussed how while sound is useful, it can also become annoying if not designed well. However, they also predicted that visual warnings (taking 1.8 – 3.6 seconds, according to the paper) will eventually be better than auditory (taking about 5 seconds, according to the paper) ones, but in chapter 5 of “Why We Make Mistakes,” the author explains that a “single two-second glance doubles the risk of an accident,” which could defeat the whole purpose of the system.

As far as future work goes, the researchers mentioned using a head-down display rather than having the display fill the entire windshield. They could also include traffic in their experiments and more types of hazards.