Wait! Before you accept the latest friend request on Facebook, are you sure your new friend is a human?
Increasingly, social media are being populated by socialbots, automated profiles that can friend, tweet, like, and comment. These socialbots are built to appear to be human. That means that, unlike 'bots such as Siri or Cleverbot, which just sound human, but are obviously not, socialbots are built to fool people in a contemporary social media Turing Test.
These bots are possible because we supply massive amounts of digitized statements in social media. There's a huge database built from encoded tweets and likes that make users' states of mind available for analysis. These states of mind can be encoded into an artificial social media profile that appears to be human, asking and answering questions and gaining followers.
To be certain, these 'bots tell us something about the state of software engineering within social media. No doubt their makers are skilled programmers, able to anticipate many social situations and make 'bots that are capable of adjusting to trends on Twitter and Facebook.
But more importantly, socialbots tell us something about ourselves. They reflect us, at least the particular images of us that appear within social media: We are liking and tweeting machines. This makes us a ripe target for "weak artificial intelligence."
What I mean by this is that socialbots are built to work within highly constrained environments where only a limited number of tasks need be done. This is so-called "weak" AI. Examples of this form of AI include video game AI (think of how Elites hide behind rocks in Halo); Sgt. Star, the U.S. Army recruiting chatbot; Siri on the iPhone and Cortana on Windows Phone; and Google self-driving cars.
"Weak" AI has been incredibly successful, in contrast to what Jerry Kaplan calls "science fiction AI," a form where machines have self-consciousness and the sort of intelligence we humans have. Science fiction AI, i.e. "strong" AI, may never happen, but weak AI is everywhere these days.
If socialbots are not "strong" AI (capable of conscious reaction to new and unexpected situations), if they are only possible because they work in highly constrained environments, we should be concerned about the state of social media. If we find our online humanity reduced to acts such as liking products, retweeting the latest news stories, or falling in love with "hot" profiles, and if this is so easily replicated by socialbots, we haven't set the bar very high for humanity. Whose intelligence is weak?
I'm not above all this, myself; I'm pretty certain I've been fooled by 'bots in Twitter a few times. And there's no denying social media use is quite complex, even if it can be imitated by 'bots. For example, it's hard to imagine 'bots that can handle the rich complexities of irony and sarcasm.
So let me alert us to the dangers of socialbots in another way. Maybe robots will be able to drive our cars or sort our socks, but I am not convinced that they should be our friends — at least, not while those who build them imagine ways that our artificial friends can manipulate us into buying useless things, divulging our desires, or toeing a political line. As I argue in my book, behind socialbots stands a massive, powerful network, one we've been hearing a lot about lately: the network of surveillance, comprised both of global corporations who buy and sell our attention and governments who demand our obedience.
Marketers, who are on one side of this surveillance network, could easily use socialbots to drive attention to brands; if you want your brand to appear to be viral, why not buy a thousand retweets? Or, if you want to learn more about your target customers, why not send an automatic envoy to befriend them and gather data on them? Similarly, as has been well-reported, politicians are using these 'bots to create the illusion of massive, grassroots support for their campaigns.
On the other side of the surveillance coin, governments are already using bots to quell dissent. A message repeated by a network of socialbots in Twitter — say, "I think the Leader is doing a fine job" — could subtly alter online debate about government policies and actions.
In the end, I can think of ways in which robots can be among us — indeed, they are in many areas of our lives — but given the combination of the surveillance state and crafty robots who pretend to be human, I can only come to one conclusion: Ban the socialbots.
A socialbot ban would likely be best enforced by the Federal Trade Commission, the agency tasked with regulating advertising. Imagine a Facebook profile that presents itself as an empathetic friend, someone to act as a sounding board for your deepest-felt anxieties about your health, but is really a bot run by the advertising firm BlueKai on the behalf of the pharmaceutical company Pfizer. This wouldn't just be false advertising; it would be deplorable manipulation, a well-programmed deceit.
A ban could also be implemented by the Federal Communications Commission as part of their regulation of political advertising, since socialbots could be used to create a false — and untraceable — sense of popular support for any politician willing to shell out the money for a socialbot network. Any artificial intelligence agents ought to be openly presented as robots paid for by the corporations and politicians who deploy them, not as false, friendly profiles built to insinuate themselves into our social networks.
Of course, this won't do much to stop federal agencies such as the FBI from deploying socialbots in a kind of COINTELPRO for our times. Activists, be warned — the next informant to infiltrate your group could be a robot. Dealing with such state-based surveillance will take more than policies from the FTC or FCC; it will take a much-needed, concerted political backlash against our surveillance state.
When it comes to life with robots, openness should be valued. If we want to divulge our innermost secrets to a robot — and if we know the other to be a robot — that's fine. But for a program to pass itself off as human to befriend and manipulate us is disconcerting, to say the least.
If you're a bot, you should announce it to the world, or face deletion. Out, damned socialbots.