Another nice idea... a heuristic that approximates the "shame value" of files. This heuristic is then used to bias the random selection of files. Before each round, the player can either accept the file offered - or in the event the file is among the worst imaginable, he can choose a "?" option that is guaranteed to be a different file... chosen with even heavier heuristic bias.
There's no way to calculate the "shame value" of a random file.
Hell, I have a text file on my computer that when opened in notepad (i.e. fixed width font) you get some nice porn (at least, what's doable with black and white ascii characters). Unless the filename was something incriminating and not generic (I don't recall offhand, I think it's like "asciitest" or something) it's shame value would be really, really low.
Likewise I have a porn story written in German on my desktop right now (I keep meaning to run it through Google translator and see if I can at least read it, I have no idea if its even
good, translation errors aside). What's that file's shame value?
And that's not even taking into consideration that I might not care if my raunchy porn collection gets shared out to the internet at large.