I found some discussion of this issue in a real-life context (people
failing to do fizzbuzz in programming interviews):
> 5 years ago
> In my experience interviewing people, it's not usually that they can't
> solve the problem. It's that they get stuck on how to solve it
> "correctly." They want to solve it the way that I expect them to, when
> really I want them to do it however the fuck they want. The result of
> FizzBuzz should never be a yes/no decision even if they get it wrong.
> The idea is to see how they think.