This is an edited version of an article that appeared in DemocracySOS, a newsletter about protecting, improving and expanding democracy and representative government, sponsored by FairVote and available at https://democracysos.substack.com/

Thinking back over the history of presidential elections, certain episodes have become the stuff of legend, impacted by unique factors that may have swung a close election. Often, those episodes tracked the development of mass communications technologies and their application in politics and campaigns.

First we flash back to 1858 and the Lincoln-Douglas debates that took place two years before the presidential election of 1860. Abraham Lincoln challenged Illinois Senator Stephen Douglas for his seat and the candidates faced off in a series of seven debates spread across several months and a number of Illinois towns. Each debate lasted three hours, and by all accounts they were electrifying and artful, two master orators at the top of their game, with thousands of people in attendance. They focused mostly on the hot-button issues of the day – slavery, states’ rights and the Dred Scott decision – but it was the new technologies of the telegraph, railroads and Pitman shorthand (phonography) that turned a local debate into a national spectacle.

The large Chicago newspapers sent phonographers using shorthand who recorded complete texts of each debate, and halfway through the debate runners were handed the phonographers’ notes. They raced for the next train to Chicago, handing them to stenographers who during the journey converted the shorthand into words, producing a transcript ready for the typesetter and for the telegrapher, who sent it out to the rest of the country after it arrived in Chicago. The Chicago newspapers published the speeches in full, sometimes within hours, and newspapers all across the country reprinted the text.

The debates effectively became national events: it was as if the two candidates were talking to the entire country. Later the debates were republished as pamphlets and the publicity made Lincoln a national figure, laying the groundwork for his successful 1860 presidential campaign in which he faced off against Douglas once again.

A few generations later, President Franklin Delano Roosevelt’s “fireside chats” used radio effectively to soothe Americans in the midst of an economic depression. Over 80 per cent of the adult radio audience tuned in, exceeding the audience for Jack Benny, Bob Hope, Fibber McGee and Molly or Amos ’n’ Andy. Novelist Saul Bellow recalls walking down the street while Roosevelt was speaking. Through lit windows families could be seen huddling at their kitchen tables or gathered in their parlors listening to the radio. Drivers had pulled over and turned on their radios to listen. “Everywhere the same voice. You could follow without missing a single word as you strolled by,” said Bellow.

Fast forward to 1960, when 70 million people watched the first televised debate in a presidential election, between Senator John F. Kennedy and Vice President Richard Nixon. Going into the debate Nixon was a six-point favourite to win. Legend has it that undecided voters who heard the debate on the radio thought that Nixon won, while those who saw it on TV and viewed Nixon’s twitchy performance thought that the photogenic Kennedy won. Post-debate polls showed Kennedy taking the lead, and he went on to win one of the closest presidential elections in U.S. history. At that moment the cosmetic theatricality of TV arrived on the national scene to become the dominant factor in political campaigns.

More recently, there have been influential television media campaigns, such as the notorious Willie Horton TV ads in 1988, which featured a convicted Black murderer in Massachusetts who raped a White woman while on prison furlough. Vice President George H. W. Bush’s campaign blasted those ads to the public in the presidential race, causing the double-digit lead of his Democratic opponent, Massachusetts Governor Michael Dukakis, to evaporate. Those TV ads pioneered a new and riveting use of multimedia broadcast technology to target specific audiences. The Bush campaign used coded words and symbolic gestures directed at White people in the south, particularly White men, to paint the Democratic Party as the party of racial minorities and decaying cities.

GOP politicians have been copying and recopying Bush’s tactics ever since. Donald Trump’s entire political career is basically one long Willie Horton ad, always provoking White voters to overreact out of their fears of you-know-who, whether it’s Black Lives Matter or, more recently, the immigrant “horde.”

Often, these sorts of game-changing episodes have been the byproduct of new forms of communication technology. Now comes the latest twist in this trajectory: the mushrooming influence of digital media platforms like Facebook, Twitter and YouTube and, even more ominously, the introduction of AI-infused “deepfake” technology.

In the 2016 presidential election, Trump’s campaign was able to use the still-new marketing tools of Facebook, Twitter and YouTube to know which voters in a handful of battleground states needed persuading so that it could sell them a candidate whom a majority of Americans did not want. To understand the frightening power of these new communication technologies in today’s divisive political climate, it’s necessary to recognize their roots in a business strategy known as “long tail marketing.”

The term long tail was first coined by tech journalist Chris Anderson in his New York Times bestselling 2006 book The Long Tail: Why the Future of Business is Selling Less of More. Its blueprint is captured by figure 1, which illustrates the difference between selling a high volume of a small handful of popular items (the head portion of the graph in yellow) versus a low volume of many more items that aren’t as popular (the longer blue tail). If you can reach those customers, the many less popular niche items making up the “long tail” sell more total items than the few most popular.

Figure 1: The long tail

Anderson developed the theory that, because of the power and reach of the internet and its ability to mass-market, many companies and their consumers are able to increasingly shift away from a business model based on everyone chasing a smaller number of popular “hit” products to one of selling individualized niche products. Products that are in low demand, or that have lower sales in terms of volume, can potentially exceed the bestsellers as long as the channel of distribution is large enough and unrestricted.

Amazon, eBay, Alibaba and a host of other online e-commerce companies have implemented the long tail strategy successfully, allowing them to realize significant profit out of selling small volumes of hard-to-find items to many customers instead of only selling large volumes of popular items. Similarly, Netflix and Amazon Prime generate a lot of revenue from renting small numbers of many less popular movies to a wide range of customers. The consequence is that Amazon reportedly sells an incredible 350 million products, a figure as unfathomable as the vastness of interstellar space.

Facebook, YouTube and Twitter have brought the concept of long tail marketing to their own advertising business model, including ads and communications used in politics and political campaigns. Their sophisticated advertising machines use precise microtargeting to show slightly different versions of the same ads to millions of niche users, based on the psychographic profile of each individual viewer. The engagement algorithms decide which content is featured at the top of users’ news feeds, and what is promoted and amplified. Virtually no two people see the same ad, as each is tweaked by autopilot via Facebook’s automated advertising system, in which algorithms perform the essential duties of ad selection (and to some extent composition).

These unprecedented “tools of virality” are being deployed by extremists of all stripes, including a small number of supercharged political actors with huge numbers of followers. Donald Trump has at his fingertips the capacity not only to reach a historically unprecedented audience size but also to long-tail-microtarget relatively small niches of users (like swing voters or the most partisan base voters) with his warped version of reality.

In his 2016 campaign for president, Trump ran 5.9 million different versions of the same basic ads, each one tweaked for individual targets, according to an internal paper written by a Facebook analyst. Each ad was rapidly tested on a unique niche of prospective voters to see which ones generated the most effective engagements. After a few ads became the runaway winners, those would be widely disseminated by algorithmically identifying more people who shared similar characteristics with the original targets.

The Trump campaign ran up to 100,000 iterations of an ad in a single day, in which language and visuals were tweaked to entice as many people as possible to click. Facebook’s internal analysis showed in detail how Trump’s campaign was better than Hillary Clinton’s at leveraging various advertising tools, with names like Custom Audiences and Lookalike Audiences,. The long tail harvesting of specific psychographic profiles was deployed to bombard millions of prospective voters with manipulative ads.

“Both campaigns spent heavily on Facebook between June and November of 2016,” according to the Facebook paper, “but Trump’s FB campaigns better leveraged FB’s ability to optimize for outcomes.” In a presidential election that was decided by only 78,000 voters across three states (Pennsylvania, Wisconsin and Michigan), the Trump campaign’s strategy of using long tail marketing of targeted ads to tens of millions of Americans not only may have decided the outcome but also certainly points the way toward the future of political campaigns.

What a difference eight years can make. The communications technologies have morphed and advanced in ways that were scarcely imaginable in 2016. Sprinkle in more AI and deepfake counterfeits and you have a potentially disastrous 2024 election now taking shape.

The new deepfake technologies are spreading like wildfire, able to create astonishingly convincing virtual identities of famous people. Last October, actor Tom Hanks issued a warning to his 9.5 million followers on Instagram that an advertisement for a dental plan using his likeness and voice substitute was fraudulent and based on an AI version of him. Then Gayle King, the nationally known cohost of CBS Mornings, warned her followers on social media that video ads using her AI likenesses to promote a weight loss “secret” were unauthorized and fake. The voices of the late Anthony Bourdain and the late Andy Warhol have both been recreated for recent documentaries using AI, and the enhanced voice of the late John Lennon was used by his surviving bandmates to produce the latest Beatles hit.

This ability of tech specialists to take digital 1s and 0s and turn them into a mimicked face, voice and video character has grown shockingly fast. Companies like Deep Voodoo are working with movie studios to put famous actors’ faces on stunt doubles and even to revive deceased movie stars. The technology has progressed so quickly that any famous person’s likeness can be replaced by a computer-generated stand-in in a matter of hours.

Now it’s being applied to politics. Millions have viewed a deepfake video manipulated to present a dead-ringer version of former President Barack Obama calling Donald Trump a dip shit. This year during the New Hampshire primary, Republican operatives paid a Texas-based company to use AI technologies to mimic the voice of Joe Biden in a robo phone call in which the fake Biden discouraged voters from voting. The caller ID was falsified to make it look as if the calls came from a former chairwoman of the New Hampshire Democratic Party.

More recently, in a harbinger of what is to come, a pivotal election in Slovakia was disrupted by an audio recording spread over digital media just days before the election, purporting to show one of the top candidates boasting about how he had rigged the election. And in another recording, this candidate’s voice was seemingly talking about raising the cost of beer (not a smart move days before an election). These AI-generated recordings went viral on digital media and left little time for outing the fakery. The targeted candidate, who was pro-NATO and sympathetic to Western interests, lost to an opponent who supported closer ties with Russian President Vladimir Putin and Moscow. Slovakian researchers believe that the AI-generated deepfakes were the work of the Russian government, though nobody knows how many votes fhey actually swayed.

The point of these technology-driven interventions is to spread mass confusion among voters in a way not previously seen. The latest communication technologies are only getting more powerful and insidious. What happens when people can’t believe their own eyes and ears about what they see and hear? These increasingly alarming features are having a disproportionate impact on our political discourse and electoral outcomes. We are in the process of taking the usual rough and tumble of politics – of “war by other means” – and stirring in increasingly manipulative communication technologies which facilitate the viral spread and amplification of sensationalized misinformation, deepfakes and conspiracy scaremongering at the flick of a computer mouse.

With the steroids of targeted deepfakes and manipulative social media now stomping around the political landscape like a Frankenstein monster out of control, how could a “winner take all” political ecosystem, in which one side in an election wins representation and all other sides lose, not foster toxic division and acrimony?