She makes a very good point that an algorithm that has bad big-O behavior can be better for humans than an algorithm with better.

For instance, I found insertion sort to be the most effective at sorting papers when I was grading. . . at least, as long as the students bothered writing their names on their homework.