Pardon the sensationalist headline of that article:
Mark says that "one thing that might explain the higher numbers here—in case people question my methods—is that I used a tarantula." Apparently, people seemed pretty eager about hitting a spider. "If you take that out it goes to 2.8% which is closer to the other turtle vs. snake studies I ended up finding."
It is still quite a surprisingly high number. At least compared to a 2008 study using the Psychopathy Checklist, which discovered that 1.2 percent of the US population were potential psychopaths. 1.2 vs 2.8 is a huge difference.
I was not aware of the other turtle and snake studies.
Note that with turtle this is the lower bound on percentage of evil; a perfectly amoral person that could e.g. kill for modest and unimportant sum of money or any other reason would still have no incentive to steer to drive over a turtle; and a significant percentage of people would simply fail to notice the turtle entirely.
This gives interesting prior for mental model of other people. Even at couple percent, psychopathy is much more common than notable intelligence or many other situations considered 'rare' or 'unlikely'. It appears to me that due to the politeness and the necessary good-until-proven-evil strategy, many people act as if they have an incredibly low prior for psychopathy, which permits easy exploitation by psychopaths. There may also be signaling reasons for pretending to have very low prior for psychopathy as one of the groups of people with high prior for psychopathy is psychopaths themselves; pretending easily becomes too natural, though.
Perhaps adjusting the priors could improve personal safety and robustness with regards to various forms of exploitation, whenever the priors are set incorrectly.