The crucial point in autonomous car adaption ought to be whether they are better than humans or not. So if they hit fewer children than human drivers do, they’re better, even if they were a further 20% better at avoiding adults.