Except, of course, this is almost certainly a machine learning algorithm. Who knows what it's actually looking for.
I'm guessing a 2d matrix and then a 1d matrix before single boolean value. :p
Every index lookup corresponds to one dimension, right? So if you're saying that function has signature "dirty(a, b, c) -> bool". But if you don't know how many bits of information a, b or c has, how does it matter?
GP posted an example of a decision tree, which is — very broadly speaking — a type of ML algorithm.
They are introspectable! You could find out. Well, Facebook could.