Bracing for Inevitable Manipulation

We really haven’t been hiding under a rock, its just the stack of reading all of us have to catch up on while so much is going on has become an archeological project here — well, OK, namely for myself.  And so my comments below about news from earlier this summer regarding Facebook’s manipulating people’s news feeds and some commentary about Facebook’s “I Voted!” button might seem like I’m really behind, but actually it dawned on me this past weekend that an important piece of our work backed by the Knight Foundation has a role in this… where “this” is actually about big data.

Facebook’s use of happy and sad words to research how they affect the mood of people’s news feeds was well covered in the news a couple of months ago.  This “emotional contagion experiment,” raised all sorts of ethical questions about research on subjects who don’t know they’re being tested.

But in the tech and political worlds, people have been equally disturbed about another kind of possible Facebook manipulation—its use, or non-use, of what it calls the “Voter Megaphone.”

What’s the Facebook Voter Megaphone?

The megaphone is the “I Voted!” button that Facebook placed on the top of News Feeds for all U.S. users over age 18 on Election Day in 2010 and 2012. Users could click the button to show their friends they voted and subsequently see which of their friends clicked the button as well.

A study, commissioned by Facebook researchers and published in the journal Nature in September 2012, determined that the “I Voted” button may have boosted 2010 turnout at least by 60,000 voters and as much as 340,000 voters because of the social-media ripple effect. People who saw that their “close” friends had voted were more likely themselves to go out and vote, the study showed.

Across a very big country, 340,000 people may not seem that much, but for campaign consultants, micro-targeters, and turnout specialists, that’s a very big number indeed, especially for 2010, a mid-term election when turnout is always lower.

Facebook did another experiment with an “I Voted!” graphic on the 2012 Election Day. Those results have not been published yet.

So who is upset about this?

Let’s walk through why people are upset at this and why it is important for the TrustTheVote Project.

Micah L. Sifry is a co-founder and the executive editor of the Personal Democracy Forum, an organization that covers the ways technology is changing politics. He’s a pro-Democrat progressive. In a July 3 blog post he spoke to one of the Facebook researchers and pressed him as to whether the Facebook experiments with the Voter Megaphone could have actually helped President Obama in the 2012 election. The researcher’s and Sifry’s conclusion is that it could have.

Here’s how: Facebook users tend to be more female, more urban, and younger. Those are all demographic groups that skew Democratic. If Facebook used its “I Voted” experiment on a random sample of Facebook’s users, and it increased turnout, it could have benefited Barack Obama.

But Sifry wasn’t cheering for his Democratic side. He pointed out that Facebook could just as easily not have offered the “I Voted” button to certain people, or to certain people in certain states and voting districts. That could just as easily lessen turnout for one party or the other. The point is that Facebook could manipulate an election and it would be very hard to tell by outsiders. It’s stealthy manipulation.

Jonathan Zittrain, a professor of law and computer science at Harvard University calls it “digital gerrymandering” and denounced it in a post in NewRepublic.com in June: “Consider a hypothetical, hotly contested future election,” Zittrain wrote. “Suppose that Mark Zuckerberg personally favors whichever candidate you don’t like. He arranges for a voting prompt to appear within the news feeds of tens of millions of active Facebook users—but unlike in the 2010 experiment, the group that will not receive the message is not chosen at random. Rather, Zuckerberg makes use of the fact that Facebook ‘likes’ can predict political views and party affiliation, even beyond the many users who proudly advertise those affiliations directly. With that knowledge, our hypothetical Zuck chooses not to spice the feeds of users unsympathetic to his views. Such machinations then flip the outcome of our hypothetical election.”

What does this have to do with the TrustTheVote Project?

Here’s why this is relevant to the TrustTheVote Project. Presidential and congressional campaigns today are high-stakes, high-tech efforts with lots of money and sophistication behind them. They have the advantage of “big data” collected from all of their outreach and social media efforts. Facebook, Twitter, other social media platforms, also have that “big data” advantage.

Election administrators, on the other hand, don’t. Well, actually they do, but they don’t have the sophistication and money to do a lot with it. Election administrators have lots of data, historical and recent, on turnout by state, congressional district, county, even down to precinct level, and they have demographic data and vote results data going back a long time.

That’s where TrustTheVote Project’s VoteStream initiative comes in. We are developing, in our Election Results Reporting System, software that local election administrators can use that will give top level and deep granular-level data about how the vote went, indexed by many different variables, down to the precinct level. The general public, not to mention reporters and campaign consultants, could immediately spot anomalies that might be worth looking at closer to see whether there was manipulation, or perhaps just brilliant targeting.

What else will VoteStream do?

VoteStream will help local election officials have a more level playing field with the wealthy “big data” players who can use voting and voter registration data for manipulative purposes. Elections officials can use the big data, instead, to inform.

Here’s another way that VoteStream, and in particular the TrustTheVote Project’s open-source election technologies, can help combat manipulation. Most election administrators are smart enough to predict that Democrats will want to increase turnout in Democratic precincts, and the same for Republican campaigns in Republican precincts. But campaigns have become way more sophisticated in their targeting than that. And these campaigns do not warn election administrators in advance about who or where they are targeting.

A campaign today, or even an outside SuperPac, using social media and other sophisticated get-out-the-vote campaigns, could greatly increase turnout suddenly in a way that local elections officials aren’t prepared for.  Election workers at a few targeted precincts could suddenly be greeted out of nowhere with a huge turnout and have inadequate vote casting and counting machines to cope. Or, a breakdown in the old voting machines combined with an unexpected spike in turnout could suddenly make for a three-hour wait to vote. This is not far-fetched. This is in part what happened in Florida in 2012 in key districts, unexpectedly high turnout which made for long lines.

So, during my archeological content dig this weekend, this connect-the-dots exercise seemed worth sharing if only to point out that lots of our work here has some real potential to help in ways we might not immediately recognize.  File it under the doctrine of unintended windfall benefit.