The bots collect an emotion from WFF, then match it to the characters actions which then go off to Twitter’s API to find something to relay.

A couple of posts that explain this:

http://www.demographicreplicator.com/2009/10/probabilistic-narrative.html

http://www.demographicreplicator.com/2009/10/artificial-affectivities.html

Also, to make the bots ‘move around the world’ the WFF emotions we collect inform the bot where to go (matching emotions to shops, places, events.)

Originally conceived as a way to prototype characters for stories, I’m now find new uses for them in market research.

Code is open source, details are on the site, come and have a play.