[sologoub]

The article seems overly sensational, but the point is fairly interesting - people do not think about unintended consequences of delegating decision making to machines.

While the crickets' fate is unfortunate, this article made me think about self-driving cars and all the wildlife that gets run over every day. For the larger animals, say a deer, the autonomous cars seems to promise the same safety as to a pedestrian, but what about smaller animals?

When people swerve to avoid a squirrel or a bunny, they can (and do) cause accidents that result in greater damage. It would be very interesting to know how the autonomous system is programmed to act in such a scenario?

[logfromblammo]

That brings to mind a study that I saw summarized, where the researcher put a fake turtle in a road and recorded the reactions of real drivers. While some would swerve to avoid, and a select few stopped their car and got out to help it across the road, there were also some who actively tried to hit the fake turtle with their cars.

I would assume that the automated system would overwhelmingly avoid any obstacle in the road, living or nonliving, except in situations where avoidance would create a greater danger to vehicle or occupants.

A slowly moving animal that does not occupy a large fraction of the lane width, is lower than the under-body clearance on most cars, and does not suddenly change its direction would overwhelmingly prefer automated vehicles. Squirrels, on the other hand, have a tendency to freak out and try to dodge when the car gets closer. The car might adjust to avoid one, only to have it juke underneath a tire at the last second. I doubt the car would do much more than to slightly adjust speed or move laterally within the bounds of the lane it currently occupies.

Great for turtles, but squirrels and rabbits don't have quite the right behavior profile to benefit from the likely collision avoidance strategy.

[waterlesscloud]

That was one of my early thoughts about automated cars: Will we see wildlife return to some areas in greater numbers? Even a small decrease in say mountain lions in LA killed by cars might cause a large increase in breeding over a few years.

What about areas with a lot of deer? Might they become even more overpopulated? Will more starve? Will they be driven even further into settled areas?

Even small changes in the number of deaths might have notable effects.

[sologoub]

So your thoughts are that roadkill is an effective population control for deer? Not really seeing this.

For the mountain lions, that would be overall positive from my view, but of course carries the risk of attacks on pets and humans. These can be mitigated with education, such as don't let your toy dog out unsupervised at night. Lived next to a nature preserve for a couple years and had a mountain lion visit the property on a few occasions. Never had a problem. Rattlesnakes... that's a completely different story...

[maxerickson]

I looked at some numbers for Michigan. Hunters take ~400,000 deer each year, while there are ~50,000 vehicle-deer accidents (I guess the majority of the collisions kill the deer).

I haven't found a convenient source for predators. There are wolves, and coyotes, bobcats and bears. I guess bobcats, coyotes and bears mostly don't go after adults, but they can have a big impact on the number of fawns.

[waterlesscloud]

Not the overall deer population, but around the margins, on the edges of settlements. Seems like it could make a difference there.