Twitter, which already allows celebrities to communicate what they had for breakfast to an irrationally interested public, is now also a robot testing ground. A research project called Socialbot has recently shown that robots have no trouble creating, or destroying, social network communities with cleverly disguised fake tweets.
All right, so, full disclosure: I have a Twitter account. I check this Twitter account daily. I occasionally twitter my own tweets (or whatever). And I even read tweets from robots. But I’m pretty sure that none of the people I follow are robots pretending to be humans, which is what happened with the Socialbots project.Three teams competed to program autonomous pieces of twittering software with the aim of influencing the social connections of as many real humans as possible. It was an actual competition, too, for points, prestige, and eventually 500 bucks. Bots were awarded one point for every follower they got on Twitter, three points for every social response, and the teams stood to lose 15 points if Twitter ever said, “hey, you’re a robot!” and banned their account.
One team decided to use Amazon’s Mechanical Turk service, which is based on swarms of humans performing simple tasks for dirt cheap. So like, a bot would get a message on Twitter, post that message on Mechanical Turk, and ask (say) 25 different people to respond to it for (say) three cents each. The bot would then pick one response and use it, and there you go, you’ve got clueless humans talking to clueless humans.
So how well did it work? From the team:
Swarms of bots with statistically-predictable social outcomes could be built and used to actively sculpt and rewire the connections of social groups online consisting of thousands (or perhaps hundreds of thousands of users).”
And that’s pretty much exactly what they’re planning to do next:
These robot swarms are capable of disconnecting user groups too, so a fun game suggested by the project organizers would be to compete to have the first Twitter robot swarm to (say) effectively disrupt some group of activists. Fun!
And no, I don’t see any possible way that any of this could possibly go wrong. Why, do you?
“The bots collectively generated close to 250 responses, and received mutual connections from close to half of the entire target set. [In] two weeks, the bots were able to heavily shape and distort the structure of the network. This included bringing people together not originally connected, and bringing together a community of activity around the bots themselves.
“We’re going to survey and identify two sites of 5,000-person unconnected Twitter communities, and over a six-to-twelve month period use waves of bots to thread and rivet those clusters together into a directly connected social bridge between those two formerly independent groups. The bot-driven social “scaffolding” will then be dropped away, completing the bridge, with swarms of bots being launched to maintain the superstructure as needed.”