>What evidence is there that the attrition we see from car accidents is actually a net-loss for society?

What does "attrition" mean? People killed in car crashes? How could that not be a net loss? Your second sentence says "furthermore" to introduce the idea of the cost of safety measures, so that seems to imply the first sentence isn't considering the cost of safety measures.

> What does "attrition" mean? People killed in car crashes?

Yes

> How could that not be a net loss?

Is every human a net-positive to society? Obviously an objective measure is impossible to establish, but I doubt everyone is a net-gain.

How random are car-related deaths? Is this actually a problem?

In general I think society is based on the idea that everyone's life has inherent worth, and thus their rights should be protected. If we start designing systems based on the idea that some people's lives have negative worth and if they die it'll be beneficial society, that flies in the face of the idea of inherent worth. It seems to me that if society says certain people's lives have negative worth so their lives don't need to be protected because their death will benefit society, that could progress into society saying that it's ok to actively kill those people.

Yes, the US does today in fact actively kill some people with the death penalty. Though it does have a lot of checks in the process to avoid making the determination wrongly. It's a very expensive and thorough process. I'm against the death penalty because it goes against the idea of the inherent value of human life. If you live in a society where jails are very primitive and can't stop determined prisoners from escaping, then the death penalty could make sense to prevent prisoners escaping and then hurting others. We have jails that can hold people though.

By and large, people very much like to stay alive.

So, if you don't care much about Bob one way or another, but Bob and his family care about Bob, on net society cares about keeping Bob alive.

You’re right, mass use of roads for everyday transportation is incredibly inefficient in almost every way: Outrageously high construction/maintenance costs, outrageously large space requirements for roads and parking at both ends, high resource and environmental costs for the cars and driving per se, high costs in human health from a sedentary lifestyle and plenty of local pollution, high costs in human time spent driving around in cars, etc.

We’d be much better off if we re-planned the whole society to switch as many trips as possible to other methods of transportation. This would require changes in land-use laws, a shift in investment toward mass transit and pedestrian/bicycle infrastructure, etc.

Making the roads we already have safer would be a good start though.