The GMF researchers, in partnership with the nonpartisan NewsGuard service, monitored engagement with two kinds of sites. The first category is composed of popular purveyors of provably false stories, such as Newsmax and One America News Network, as well as smaller “Trojan horse” outlets posing, for example, as local papers. The second category includes outlets judged irresponsible in their information-gathering and presentation. These include Fox News, Breitbart and others whose tendency to distort the facts, rather than flat-out make things up, means their content is less likely to be rated false by Facebook’s checkers, and therefore less likely to be demoted in people’s feeds.
The results were troubling but not surprising. Shares of tweets from deceptive sites by verified accounts on Twitter reached an all-time peak of 47 million in the final quarter of last year — almost one-third of the total 155 million shares by verified accounts of links to U.S.-based sites. Facebook saw a decline during the same period in interactions with all sites, but interactions with deceptive sites increased over previous years.
This is, of course, more or less what the conventional wisdom has come to expect. These are profit-seeking corporations, and profit-making ones, too: Facebook announced a mind-boggling $11.2 billion in quarterly profits the same day the GMF report was released. Twitter rakes in far less cash but has similar incentives. Social media sites want to drive engagement, which comes more easily from the sensationalism that marks most conspiracy theories than it does from sober-minded news reporting. Yet if these companies hope to continue as money-making machines without meddling by lawmakers, they must prove conventional wisdom wrong — by showing that they can do what’s right.
We got a taste of what’s right around the November election. Facebook altered its algorithm to prop up mainstream outlets (including, in full disclosure, this one); GMF reports that interactions with its sample of credible and transparent outlets were highest with those measures in place. Twitter disabled likes and retweets on some misleading posts and forced users to take extra steps before sharing articles; after the friction was removed, misleading posts spiked. None of this was censorship, but merely changes in design that reversed the status quo by privileging the trustworthy instead of what can’t be trusted.
These interventions today are the exception, yet protecting information integrity ought to be the rule. Discovering the most effective means to achieving that end will require some trial and error. So far, however, platforms aren’t trying nearly hard enough.