Co-Founder of Geocodio here who designed the scoring system :)

I suppose you could call it inspired by Bayesian inference since we're using multiple pieces of independent evidence to calculate a score, though that makes it sound a bit fancier than it is and we aren't using the Bayes' theorem. But it's possible I had that in the back of my head from a game theory class I took long ago.

But for the fun of it, let's model it that way:

Probability (Spam | disposable email domain, IP address, etc... ) = [probability(disposable email domain, IP address, etc... | spam) x prior probability(spam rate)] / probability(disposable email domain, IP address, etc...)

Or something like that.

Also — it's a delight to have one of Patrick's articles mentioned in connection with this!