The assumption, I guess, that propels this is that once you...

The assumption, I guess, that propels this is that once you randomly test students, they'll just stop experimenting. But behavior is really complex. It doesn't always work that simply.