Artificial intelligence is capturing the public consciousness. Despite having obvious limitations, it is becoming an increasingly useful tool. It produces illustrations for my Substack, Joxley Writes. They are not exactly perfect, but good enough and effectively generated for free, making them better than stealing others’ images and cheaper than hiring a cartoonist. It’s perhaps only my own vanity that stops me from doing the same for the text.
The technology, and the speed at which it has progressed, has given rise to all sorts of existential debates about the threat it poses. Some are prosaic, about the jobs (like my prospective cartoonist) that have been lost to the self-generative capacities of computers – others more remote, such as the risk of computers becoming so advanced they kill off or enslave us. Little attention has been given so far, it seems, to the impact it could have on our politics.
Every major new technology has come with political consequences – not just in the broad social changes they have facilitated, but how ideas themselves are spread. The rise of the printing press was intrinsically linked to the spread of the Reformation. The Enlightenment thrived on cheap pamphlets and, some argue, newly imported coffee. Railroads, radio and television were successively employed to sway electoral favour, harnessed by all sides to spread their message.
In recent years, internet technologies have been key to major election victories. In what now seems like the early archaeology of the internet era, Barack Obama’s 2008 triumph was the “email election”. His campaign was hugely innovative, and successful, in using that method of communication to raise funds and motivate voters. In 2008 and 2012, his email campaigns allowed him to outraise and outspend his rivals, with his best subject lines raising multiples times what the less effective ones pulled in.
Through the 2010s, the focus of electioneering moved from email to social media. The smartest political campaigns pulled together both organic and paid-for content to spread their message. The Tories in 2015 ran a particularly deft Facebook campaign, while Vote Leave did the same a year later (notwithstanding the various controversies about how data was obtained and used). In the same year, Donald Trump showed how Twitter could be used to get out unrelentingly simple messages his opponents couldn’t bat back. By 2017, digital campaigning had become a key plank of election campaigns, and one in which Jeremy Corbyn’s Labour largely outplayed the Tories – the failure to repeat the success in 2019 was another big reason why it lost.
Political campaigning is ultimately a form of advertising, and so it is unsurprising that ad (wo)men’s ideas proliferate – or that the party that incorporates them most effectively will win the day. Vast amounts of corporate expertise have been built up in how to make people click, and how to persuade further them once they have. And the data generated by social media means your message can be much more precise than a billboard in the high street. As AI begins to proliferate in the commercial world, its practitioners will be looking for new political clients.
There are two obvious uses for the diffusion of AI into politics. The first is predictive – thinking about who might vote for you and how to motivate them. This is already done by political parties, relying on a mix of their data (that’s why they knock on your door) and commercially available sources. At its simplest, this is about identifying cohorts and appealing to them. If, for example, you find that lots of married white pensioners who own their homes in Surrey will vote for you, you use the data to predict who on the register fits that character and try to make sure they vote.
Beyond that lies a whole world of targeting opportunities. Whether advertising digitally or with leaflets, you can try to predict who will be amenable to which messages. Single mothers might be particularly attracted to an offering of childcare, and people in a certain area of a constituency will care more about a new bus route than those who never use it. The principle remains the same, but the data gets increasingly complex. Social media was the first wave of this, but AI elevates it.
The Democrats have already made use of this in America. Their 2022 midterm campaign used AI to solicit donations in a “small money” campaign that delivered almost ten times what it cost to run. AI can do work that is simply beyond the capacity of human teams of analysts, such as looking at far more variables to break voters down into smaller pockets. Instead of a few hundred identifying variables, AI can navigate thousands, adopting personalised approaches to target voters with greater precision – whether trying to get money or votes from them. While British parties are more limited in the resources they can deploy (the Democrat “small money” campaign costs about half as much as an entire UK election campaign), whoever is first to effectively deploy this sort of AI targeting will reap dividends.
The second clear use of AI in politics is its generative capacity. As much as you may want to target small groups of voters, this is historically impeded by the time it takes to actually get a message out. It’s not uncommon in a tight marginal to put out a leaflet targeting a few streets – but doing so takes a few hours of writing and editing messaging. AI cuts that to seconds. Playing around with ChatGPT in preparation for this blog, I could write a full election address in about two minutes from a one-sentence prompt. With a more honed command, you could personalise messages to dozens of sub-sections of voters in the time it took to write one constituency-level address.
[See also: Only philosophy can beat AI]
By providing both data analysis and text and image generation, AI is likely to usher in an era of hyper-personalised election campaigning. The same old methods, of finding your potential supporters and appealing to them will become far easier and more precise. Done well, AI campaigning will limit waste on pursuing people who are unlikely to support you with messages they don’t like.
In any individual election, this offers a big advantage to the party that gets it right. Over the past decade and a half, in both the US and the UK we have seen the impact of the best digital strategy become election-deciding. The same will be true of AI. However, it also opens up new opportunities for emerging parties, making them less reliant on the human labour of volunteers. Cutting out the work involved in these areas nullifies one of the advantages of having lots of people already signed up for you.
The rise of AI electioneering also adds to the difficulty of regulating elections. Most of our electoral law is still rooted in the era of placards and pamphlets, and struggles to deal with the realities of digital campaigning. Election adverts are subject to strict rules about truthfulness and declaring their provenance (the imprint, saying who published them is required) – which becomes harder to track as micro-targeting means an individual ad can be seen by only a handful of people. The proliferation of AI-generated content will increase this difficulty, risking the regulators being overtaken by technology.
Indeed, AI also increases the risk of influence from illegitimate actors. The social media elections have been beset by allegations of malpractice, some proved, some overblown – but the same tools that facilitate political campaigns could be adopted by malicious actors. Though the term (ironically enough) has been warped and twisted, “fake news” was a real issue in the 2016 campaign, as websites designed to look like legitimate sources infiltrated social media with made-up stories – such as the Pope’s endorsement of Trump. AI makes this sort of thing child’s play.
I got a bot to write the above article that an MP had been arrested. It would be trivially straightforward to make this look like it was lifted from a legitimate news site and circulate it on WhatsApp or social media. Other AIs make it easy to generate photos of the event and package it into a narrative many will believe – the era of fake news could become supercharged. Beyond elections, the same technology could be used to create incidents that stoke community or religious tension, with a small network of people able to generate thousands of stories and images and make them look convincing in a very short period of time.
There are philosophers and computer scientists who can debate at length what AI means for our society and the wider impacts it could have. But there are at least two obvious use cases for it on the horizon of politics. The next election, or perhaps the one after that, could well be decided by the party that utilises it best. With it comes risks that the bad actors – whether foreign powers, vested interests or mere miscreants – could seize on the same things with disastrous consequences. Parties, regulators and the public have already struggled with the rise of campaigning technologies. We’ve moved past the email election and the social media election – the AI election is coming.
John Oxley is a regular contributor to the New Statesman. A version of this piece originally appeared on his Substack “Joxley Writes”.
[See also: Big Tech’s race for our data is on]