September 27, 2020

Why Facebook’s political-ad ban is taking on the wrong problem

When Mark Zuckerberg announced that Facebook would stop accepting political advertising in the week before the US presidential...

When Mark Zuckerberg announced that Facebook would stop accepting political advertising in the week before the US presidential election, he was responding to widespread fear that social media has outsize power to change the balance of an election.  

Political campaigns have long believed that direct voter contact and personalized messaging are effective tools to convince people to vote for a particular candidate. But in 2016, it seemed that social media was amplifying this threat, and that invasive data-gathering and sophisticated political targeting had suddenly created a recipe for democratic disaster. 

The idea of algorithmic manipulation schemes brainwashing large swaths of the US electorate online is a nice way to explain the polarized nature of American public opinion. But experts say it’s actually pretty unlikely that targeted political advertising has had much influence on voter behavior at all.  

“Very quickly you get absolutely nowhere” 

Much of the reasoning behind the ban relies on the idea that social media can convince undecided voters. This has been the narrative since the 2016 election, when Cambridge Analytica claimed it used “psychological warfare” to manipulate vulnerable undecided voters on Facebook into believing fake news and convincing them to vote for Donald Trump. The Guardian reported extensively on the Cambridge Analytica’s idea “to bring big data and social media to an established military methodology—‘information operations’—then turn it on the US electorate.”  

But in reality, campaigns still can’t persuade undecided voters much better than they could 10 years ago. 

Some suggest that associating certain online attributes with voter profiles allows campaigns to group target voters into smaller, more specific groups that care about particular things, which might offer an avenue to getting them to vote a certain way. For example, you could assume all independent first-time Minnesota voters who have liked the Bass Pro Shop are likely to care about gun rights.  

But Eitan Hersh, an associate professor at Tufts University, says these assumptions get layered with errors. A campaign might assume that “the person who watches Jersey Shore has X kind of personality traits,” he says, but “those things aren’t going to be perfectly correct.”  
 
“Then I’m going to try to make an ad that is focused on that personality trait. Go to any ad seller: how easy is that to make an ad just right for that personality trait? And then it has to come at exactly the right moment on your timeline where you’re receptive to it. When you add all of these layers of error atop each other, very quickly, you get absolutely nowhere. It’s just all noise.” 

Even if these errors didn’t exist, it’s nearly impossible to measure whether ads were effective in changing somebody’s voting behavior. Voting, after all, is secret. 

That doesn’t mean advertising can’t be effective, however. In fact, the online targeted political advertising system has advanced in two meaningful ways: first, it has allowed campaigns to more accurately sort decided and undecided voters using data, and second, messaging has gotten more effective as a result of sophisticated A/B testing. 

The bigger problem 

But the true strength of online political advertising has been in sowing discord. Social-media networks function by running powerful content recommendation algorithms that are known to put people in echo chambers of narrow information and have at times been gamed by powerful actors. Instead of getting voters to switch their position, political messages delivered this way are actually much more effective at fragmenting public opinion. They don’t persuade voters to change their behavior as much as they reinforce the beliefs of already-decided voters, often pushing them into a more extreme position than before. That means the ads being banned—the ones from the campaigns—are not what is changing democracy; it’s the recommendation algorithms themselves that increase the polarization and decrease the civility of the electorate. 

Sam Woolley, the project director for propaganda research at the Center for Media Engagement at the University of Texas, says that while he’s “glad that Facebook is making moves to get rid of political ads,” he wonders “to what extent the social-media firms are going to continue to take small steps when they really need to be addressing a problem that is ecosystem-wide.”  

“Political ads are just the tip of the iceberg,” he says. “Social media has horrendously exacerbated polarization and splintering because it has allowed people to become more siloed and less civil because they’re not engaging as much in face-to-face communications, because they’re behind a wall of anonymity and because they don’t really see consequences for the things they do.” These algorithms may seem mathematical and objective, but Woolley says the system is “incredibly subjective,” with many human decisions behind how and why particular content gets recommended. 

So Facebook’s ban ahead of November 3 won’t do much to change voter behavior. Indeed, since Facebook’s algorithms give more weight to posts with some time and circulation behind them, Zuckerberg’s ban might not have any significant impact at all. 

Tackling the rest of the iceberg requires a total reframing of what social-media networks actually are. 

“There’s no denying that the fundamental alteration of our media system from broadcast to social media has irreparably changed the way we share information, and also the ways in which we form opinions, and also the ways in which we get along—or don’t get along,” he says. 

What does this mean for democracy? 

This is not an entirely new problem. The American political system has used targeted political advertising for decades, long before the internet. In the 1950s, before cookies tracked your online behavior to create detailed logs, campaigns would send canvassers to specific addresses that were home to undecided voters. In the 1960s, before online advertisers started serving custom-made ads that convinced you your iPhone was listening to your conversations, data scientists were engineering messages aimed at small groups of persuadable voters. 

Social media’s role has not been to dramatically change the direction of this system, but to intensify the polarization and fragmentation it causes. On top of this, larger and more extreme groups also become vectors of misinformation and propaganda, which accelerates and worsens the problem. These challenges go far beyond Facebook’s ban—they challenge the whole online economic and information ecosystem. 

“Social-media networks, in particular, have challenged what we think of as democracy,” says Woolley. “They’ve undermined our democratic communication system in a big way, contrary to what we thought they were going to do. That being said, I do believe that democracy is a work in progress.” 

Source link