There are over 10,000 collisions between vehicles and wildlife in Ontario every year. Most are not fatal for those in the vehicle, but the costs associated with such a collision in Canada total about $800 million annually.
While automotive safety systems already warn drivers of some hazards – such as a vehicle in their blind spot – identifying unexpected hazards, including potential collisions with wildlife, in dynamic road situations is a more complicated task.
Now, cognitive psychologists from the University of Toronto have found that simple safety alerts may work just as well as complicated systems, which are more vulnerable to errors, when it comes to alerting drivers of impending wildlife collisions.
In a recent publication in Cognitive Research: Principles and Implications, Anna Kosovicheva and Benjamin Wolfe – co-directors of the Applied Perception and Psychophysics Laboratory (APPLY) at U of T Mississauga – examined ways to make vehicle safety alerts more effective. To conduct their analysis, the researchers used dashcam footage of actual driving situations they obtained from YouTube. Using methods that draw upon fundamental cognitive psychology research, they tested three different types of attentional cues that warned research participant, or driver, of a looming hazard, and evaluated how each one affected their response time.
One warning drew a driver’s attention directly to the hazard. Called a spatiotemporal valid cue, this accurately superimposed a graphic of expanding red rings around the hazard itself – as though the safety system had identified the hazard in the correct place at the correct time. This worked. Driver response time was about 60 milliseconds faster.
“If you are not a vision scientist used to thinking in milliseconds, that might not sound like a lot,” says Wolfe, an assistant professor in the department of psychology at U of T Mississauga. “But if you are driving on the highway, it could be two or three metres. That is not enough to brake fully, but could be enough to swerve and avoid a collision.”
But when a hazard was inaccurately identified, it had the opposite effect. Drivers were also given a type of warning called a spatiotemporal invalid cue. This inaccurately superimposed the set of expanding red rings around some other object in the scene. That drew the driver’s attention away from the hazard and slowed response time by about 60 milliseconds.
“This is fairly disconcerting, from a road safety standpoint. No automated vehicle systems will ever be perfect,” says Wolfe. “The engineers building them will get it right most of the time, but sometimes systems will fail.”
Kosovicheva and Wolfe also tested a third type of cue that is simpler to execute, from an engineering standpoint. Called a temporal valid cue, drivers were warned of the presence of a hazard by a red bar at the bottom of the screen. This cue came at the right time but it did not identify where the hazard was located. Still, it had a roughly equivalent effect to the cue that had accurately pointed to it. Response times were improved by about 60 milliseconds.
“This suggests that while complicated engineering solutions can be effective, simple alerts can be effective too,” says Kosovicheva, who is also an assistant professor of psychology at U of T Mississauga. “Just having the information in time can be helpful, and if you are going to have a spatial component to a safety alert, the information about where a hazard is needs to be really accurate.”