A student who has excess bandwidth after acing a test is more likely to want to poke and explore the limitations of that system than a student who, at capacity, is almost perfect.

Unscientific, but I'd bet there are two major cheating populations: 95%+, and something like 50-75%. By definition there are a lot more of the mid-tier-barely-getting-by cheaters, but they also aren't as smart and are only dangerous to societal trust in aggregate.

Come to think of it, I missed the third major population: pre-meds