“He’s not human. He’s a robot. Dante Pryor is a socialbot,” lawyer and series’ heroine Alicia Florrick says in the latest episode of the TV series The Good Wife.
“How could a robot defame your client?” the judge character asks her.
“It's designed to repackage comments and gossip by others and disperse it onto sites,” Florrick responds. “It's a computerized version of the worst part of human nature.”
These so-called socialbots, increasingly part of our Zeitgeist, are more or less the same in real life. Anna Jobin, a researcher at the Swiss Federal Institute of Technology and at Lausanne University who is currently working on a doctoral thesis about our interactions with algorithms, notes that socialbots are ubiquitous in political campaigns, for example.
“An algorithm creates thousands of socialbots to exert influence or to drown divergent opinions,” she says.
The Mexicans, Russians and Syrians all have been accused of deploying armies of socialbots over the last three years. And Indiana University’s research group Truthy observed the use of this electoral weapon during U.S. congressional elections in 2010.
A socialbot is basically an automated computer program capable of creating profiles on social networks, complete with names and photos, before establishing connections with users, chatting, commenting on friends’ posts, and fueling discussions. It picks up on the living material uploaded by humans: attractive faces taken from the website Hotornot.com, elements of identity found on social networks, and clear-cut opinions gathered in the news or in tweets.
Websites such as 10minute-mail.com allow the swarm of “robots” to obtain disposable email addresses to sign up to networks. Other online services are used to get around the CAPTCHA obstacle — the sequence of bent letters and numbers a user must decode to “prove that you are human.” The rest, the formula that mixes a variety of data to simulate the behavior of an average person, is a matter of algorithms.
Most times, socialbots land on social networks as invaders. Their aims are political or commercial and are treated as unwelcome when discovered. But sometimes, the same site where they are active also creates them. “It’s the case for some dating sites,” Jobin says. “When there are more men than women among the users, the website may then deploy a large number of fake female profiles created by algorithms to increase the activity.”
She says that a single program can create up to 20,000 profiles. “These are malicious and deceptive manipulations,” Jobin says.
Hide and seek
Internet robots (or bots) are nothing new. When Amazon suggests purchases, it’s because an algorithm analyzed your previous purchases to determine your preferences. There is an attempt to influence your purchases, yes, but no fraud per se when algorithms are used in this way. But socialbots are, as the saying goes, a different kettle of fish.
“The website Netflix — an online video-on-demand service — has determined that in 60% of cases, consumers choose according to suggestions based on the history of their purchases,” Jobin says.
And that leaves the other 40%, those whose viewing choices wander off the beaten track, as targets. The question is whether it’s possible to steer these users as well? “Start-ups in the Silicon Valley are working on it,” Jobin says. “They talk about "serendipity algorithms'. In the case of socialbots, in a similar way, they create inconsistencies, irregularities, randomness, so their behavior seems more credible.”
It’s not just about deceiving users, but also the social network itself. “On the one hand, there are algorithms conceived to influence users without being discovered,” Jobin says. “On the other, there are algorithms made to detect these social algorithms. It’s a game of hide and seek.”
When the herd of socialbots and their human owners (the bot herders, or “robot keepers”) succeed, no one talks about it because their activity is undetected. The socialbots that make the most headlines are those that are created with the singular aim to show how a socialbot works.
Last summer, a Brazilian bot called Carina Santos managed to become a journalist, influencing Twitter by automatically recycling tweets from the newspaper Globo.
“Its creators — researchers at the Ouro Preto Federal University — wanted to show that the way Twitter measures influence is not reliable,” Jobin says.