Feature idea: a simple way to automatically group conversations into "probably a human" vs. "probably a bot"
The most interesting spa.mnesty threads are, of course, the ones where there is a human on the other end. The bot vs. bot conversations tend to not be very interesting.
Here's a simple filter that could automatically separate the bot threads from the human ones with fairly high confidence.
Simply scan the spammer's replies and look near the top for the first or last name of the "mnesty.com employee" who replied to the spam. If either name appears near the top, it's probably a human reply. If it doesn't, it's probably a bot.