On Human Irrationality. The risk of a small meteor hitting the earth is pretty small, but the effects could be devastating. Gregg Easterbrook writes:

In 1908, an object 250 feet across hit Tunguska, Siberia, flattening trees for 1,000 square miles and detonating with a force estimated at 10 megatons, or 700 times the power of the Hiroshima blast. Had the Tunguska rock hit Moscow or Tokyo, those cities would have been seared out of existence. In 1490, an estimated 10,000 people were killed when a mid-sized meteorite hit China. In the year 535, a series of mid-sized meteorite strikes around the globe kicked enough dust and debris into the atmosphere to cause several years of cruel winters, helping push Europe into the Dark Ages. Ten thousand years ago, just as modern Homo sapiens were making the first attempts at controlled agriculture, something enormous struck the Argentine Pampas, obliterating a significant chunk of the South American ecology with a force thought to be 18,000 times that of the Hiroshima bomb...

A few weeks ago, researchers at the Harvard-Smithsonian Center for Astrophysics briefly believed that an asteroid packing the power of a one-megaton nuclear bomb was days away from striking North America. They turned out to be wrong, needless to say. Had they been right and the president informed that a killer rock from space was approaching, when he asked what could be done, he would have been told, "Sir, we can do absolutely nothing."

Easterbrook proposes that NASA make the prevention of asteriod collisions part of its mission. He argues that NASA has nothing to do anyway and ends up inventing missions for itself in the form of the International Space Station and the Moon-based Mars mission initiative.

He has a good point. But, of course, this will never happen. If we were rational creatures, we'd multiply the probability of an asteroid collision by the expected devastation to decide whether a proposal like this would be worthwhile; but, as it happens, we tend to simply disregard low-probability possibilities.

Evolution still has some work to do.