In the midst of a bitter and disjointed Democratic primary, Donald Trump looms large on the political horizon, threatening to replicate his performance of 2016, after emerging relatively unscathed from an endless series of scandals and attempts to impeach him. Democrats are presumably—and rightly—arguing about the political strategies required to unseat the incumbent, a historically difficult thing to do at the best of times. There is, however, a different way to frame the issue if the lessons of 2016 have been properly learned: should Facebook let Trump win again? That is, after analysing Trump’s unprecedented success in using Facebook to win the last presidential election, are Democrats better placed to take him on in the political sphere or should they attack the best tool in his re-election arsenal? Should Facebook itself take steps to limit its advertising’s reach and influence on the American political process, knowing the advantages its services afford the Trump campaign?
The world of social media advertising is notoriously nebulous: a swamp of misconceptions and poorly understood services that confound traditional campaign strategies and make specialized consultants obscenely rich. Andrew Bosworth, a Facebook executive who was in charge of the ad division during the 2016 presidential election campaign comments that Trump, “didn’t get elected because of Russia or misinformation or Cambridge Analytica. He got elected because he ran the single best digital ad campaign I’ve ever seen from any advertiser. Period.” The implication is obvious: Facebook provides an advertising platform that produces better results than traditional media, when used in the correct way. This reveals a darker truth: successful digital advertising often requires deep knowledge of the specific individuals it is targeting, and an understanding of how to most effectively exploit their hopes, biases, fears and aspirations. When the Trump campaign speaks of knowing its voters it does not mean it knows them as a collective with a set of common characteristics: it knows them by name, address, browsing history, Big 5 personality traits and attendance at official Republican events. The Trump campaign is working to persuade likely voters at an individual level. Facebook allows, even encourages, this strategy.
Digital advertising campaigns are procedural: they are heavily reliant on data and require a test-and-learn approach in order to maximize success metrics like voter registration, rally attendance and donations. Brad Parscale, the man overseeing Donald Trump’s digital advertising spend in 2016, designed a Facebook advertising strategy that was far superior to Hillary Clinton’s. As Trump’s 2020 campaign manager he has the potential to do so again, in response to whoever Trump’s opponent may be. So why can’t Democrats replicate Trump’s Facebook strategy—assuming that Mike Bloomberg follows through on his promise to bankroll the campaign of whoever wins the nomination, there will be effectively no spending limits, and no reason for Democrats not to outspend Trump in targeting voters in critical swing states.
However, elections in the United States present a special challenge to digital campaigns: as voting is not compulsory, campaigns can try to secure votes for a candidate, while also attempting to suppress the opposition. There is a strategy for gaining votes—identify supporters and motivate them to turn up on election day—running in tandem with a suppression operation, which focuses on identifying those likely to vote for another candidate and convincing them to stay home. Facebook is the major battleground for this for a few key reasons: around 70% of US adults use it regularly and a growing number of them utilize Facebook as their primary news source.
However, there is another reason why Facebook is a critical piece of the election advertising puzzle, which has increasingly raised the ire of liberal Democratic candidates, such as Elizabeth Warren: it has the ability to shape its users’ perceptions of reality, perceptions that can be guided and reinforced by sophisticated advertising operations, such as the Trump campaign. Facebook itself creates a feedback loop that is likely to concentrate the biases of any user. The genius of the Trump campaign in 2016 was to find the most contentious and emotive topics—immigration, healthcare, treatment of veterans, perceived unpatriotic behaviour by Democrats—and shape them into specific advertisements intended for users who were already interested in them. The Trump campaign was adept at nudging people towards identifying and solidifying their own beliefs.
This is why the Democratic Party probably won’t be able to use raw financial might to unseat Trump: to influence users into taking a particular course of action, you need to know about their preferences and behaviours in great detail. It is not sufficient to blast out a blanket campaign message ad infinitum: the true power of social media advertising lies in playing the long game, in crafting a highly nuanced data set about specific individuals, which can be used to understand the types of messages they find persuasive and learn from past campaigns which messages work.
The Trump campaign has been building its master voter data file since 2016. The Democratic Party’s centralised voter data will undoubtedly be less refined and useful than the Republican equivalent, and comprised of voters who disagree with the fundamental political positions of certain candidates still in the race. Trump enjoys upwards of 90% approval in the Republican party: the Democrats are yet to decide whether or not to embrace socialism, at least in a nominal sense. The division within the Democratic Party is precisely the kind of message that the Trump campaign will amplify and exploit in its voter suppression efforts. A brokered Democratic Convention, should it occur, would be an astounding win for the Trump campaign, who would view it as almost irrefutable proof that the Democratic Party is hopelessly ideologically divided and disorganised. At least, this is how it will be presented to probable Democratic voters in their Facebook newsfeeds: as yet another event that can be represented in such a way as to blur objective truth without being definitively false. Democrats will argue that a brokered convention demonstrates the unprecedented strength of the field; Republicans will say that it shows that the party is teetering on the brink of chaos. A plausible argument is all that’s required to create a decent suppression ad, not an objective analysis of all possible positions.
It is tempting to be sceptical as to how powerful any single advertising channel can be, given that many candidates spend immense sums of money on traditional media, such as television ads. But Facebook is a not a unidirectional medium: Facebook was primarily intended to be a content host, but has evolved into a powerful machine learning tool, which refines and optimises the type of content that its users will find compelling. The types of content that users seek on Facebook, as well as the material that they like can be used to build predictive personality models: the content you consume and interact with acts as a proxy for the type of person you are, providing accurate insights into your key personality characteristics. The Cambridge Analytica scandal exposed the extent to which Facebook data and third party applications could facilitate the personality profiling that precedes microtargeting—building increasingly refined sets of voters for whom highly tailored messages can be designed. The after-effects of this scandal have largely failed to account for the fact that the profiling work undertaken with scraped Facebook data is still valid, and still has immense utility for the targeting of voters for the purposes of incentivising turnout or suppression.
Brad Parscale has described the microtargeting abilities of the Trump campaign, including the way the campaign prides itself on identifying geographical groups with similar policy concerns, such as rural Ohians who believe a border wall should be built. Crucially, this is where the Trump digital strategy intersects with traditional campaign rallies. Trump will often focus on topics that are disproportionately popular—or things that invoke specific local concern—in the places that he visits. The cycle that Facebook supports is recursive: Facebook allows the Trump campaign to target known geographical voter groups with specific messaging, reinforce it at physical campaign rallies, track the attendees of the rallies and predict whether or not they’re likely to vote, all of which feeds back into refined and more effective digital messaging. The collation of data from multiple sources—Facebook activity, personality profiling, rally attendance data—into a master set allows thousands of messaging combinations to be tested in real time, and digital advertising spend to be linked to real world actions with high value, like voter registration. Facebook is not the only digital tool that could be used to achieve this, but it is currently the most powerful. On Facebook, it is even possible to find new audiences for your content after creating an initial round of profiles: the Trump campaign made use of a Facebook feature called Lookalike Audiences, which suggests individuals that may be responsive to an ad campaign, based on known results from previous advertisements. Facebook therefore provides an automated way to easily find potential Republican voters, without the imprecise and difficult process of canvassing neighbourhoods by foot or signing up volunteers for the joyless task of cold calling.
Should Facebook then be accountable for the outcomes of the tools that it provides? Facebook, to a large extent, does not seem willing to engage in any kind of introspection as to the influence it may have over the democratic process. Despite pressure from numerous fronts, Facebook has strongly indicated that it does not believe it has any right or duty to interfere in the forms of political advertising undertaken on its platform, apart from enforcing the community standards that apply to all advertisers. In this, it contrasts starkly with other platforms, such as Twitter, which has decided to ban political advertising entirely. At the core of both decisions is the same ethical issue: a platform provider can choose to support political advertising in general, including advertising that contains known misrepresentations or distortions, or it can choose to exit the game. To take a third way, moderating some but not all political content, invites the ugly controversy over political bias that has fuelled the fake news narrative in the Trump era, and the near impossible task of defining the boundaries of acceptable political positions. By taking a seemingly neutral and apolitical stance, Facebook will enormously benefit the Trump campaign. The frustration that many liberals feel in knowing that Facebook could feasibly prevent Trump’s re-election is understandable, but we should remember that all political campaigns are vying to achieve what Trump has already mastered.
The fact that Facebook has successfully created an excellent advertising and targeting tool is consistent with the objectives of a platform provider: the goal of advertising is to influence behaviour and perception, and some new platforms are far more capable of doing this than the mass media channels of the past century. The notion that Facebook be penalised for creating a tool that is far more effective than television or print media in convincing voters to take a certain course of action is antithetical to the fundamental premises of an election campaign. Digital advertising has always offered the potential to undertake the hyper-individualisation of advertising content—making it a perfect tool for political advertisers, who know the niche issues that are most important to specific voters.
That the Trump campaign may have benefited from data obtained under false pretences is now beside the point: it has been legal, and possible, to scrape the Facebook social graph via publicly accessible tools for a long time, yet few political campaigns have had the intent and ability to do so in a way that could change the course of an election. Mike Bloomberg has already spent upwards of $320 million on television advertising alone—how can we gauge the influence of incessant television coverage against the more nuanced approach of the Trump campaign? If Bloomberg is allowed to spend his way to the presidency, why shouldn’t Trump be allowed to construct an army of committed voters using the algorithms that Facebook provides? In deciding to allow advertising, Facebook—or Mark Zuckerberg—must have already grappled with this issue in detail. To disallow political advertising on Facebook at this point in the election cycle would be a proxy decision to oppose Trump, such is his dominance on the platform. Facebook has, in all likelihood, produced the best tool that campaigners have ever had the possibility of using—but very few can wield it effectively, demonstrating that not all campaigns are created equal. Facebook is the core of the issue, but it is not the issue. Democracy is an exercise in persuasion, and the time will come when Facebook is no longer the preeminent method by which to win hearts and minds.