I wonder if certain mathematical topics should automatically get extra scrutiny? While there are, for example, good questions which can be asked about the Collatz conjecture, most of the questions on MSE related to that conjecture are not very good.
On the assumption that the tool being built here can sense "levels" of bad, terms like "incompleteness", "Collatz", "Riemann hypothesis", "diagonalization", etc should perhaps add some non-zero amount of "badness" to the algorithmic understanding of post quality.
Any of those terms together with any post over a certain length is nigh certain to be crankery. :)
