Assuming ignorance by Occams Razor works when the issue it is assumed on/for matches a Gaussian distribution. Seeing as that which was fed to the bot was rather one-sided, the assumption of ignorance seems a lot less likely than the assumption of intend.
I was assuming ignorance by the bot designers of the eventual outcome, not of the people "feeding" it information.
Who knows if it were trolls, ignoramuses or people pointing out a morality flaw in a new technique.
Oh, trolls, I feel confident enough in that, but I'm not too bothered by what they thought they were doing at the time, only by what they achieved.
I believe it to be somewhat telling that apparently our first dominant instinct in this case was to flood the thing with racism and bigotry.
This may have to do with where it was "released". As alluded to earlier - what if the program was given free reign somewhere else, or not identified as to it's purpose?
If you put your nice scarf in the kitty litter tray, you can bet it's going to smell like cat piss.
(I still hold that the people who wrote the program either didn't know, didn't think, or it didn't occur to them that Twitter is the communications equivalent of a kitty litter tray - I'm fairly sure it's not worth any of their jobs to intentionally make microsoft look foolish)
I never saw the thing in operation, so all my thoughts on the matter are wildly speculative.
Anyway, the Marines have a phrase - Good Initiative, Bad Judgment - that probably applies to these sorts of endeavours.
