Monsters loom, threatening our democracies. They come in many guises, from online misinformation networks and deepfakes to social media bots and psychological microtargeting that uses our personal data to tailor political messages to our interests, attitudes and demographics.
With around half the world’s population going to the polls in 2024, democracy might seem to be in good health. Yet many worry that it is being undermined by powerful new digital technologies with the potential to target individuals, manipulate voters and sway elections. Fears of digital influence derive some of their power from the novelty of the technologies – we aren’t long into the online age and nobody fully understands what is happening now, let alone where things are going. Each new technology feels unfamiliar and can leave us sensing that the rules of the game are being rewritten. But are these concerns justified?
Advertisement
We are among a growing number of researchers with expertise in political science and psychology who are attempting to draw these monsters out of the shadows. Our studies aim to find out how the new technologies are being used and by whom, as well as how effective they are as tools of propaganda. By carefully defining the idea of digital manipulation, we now know more about the threat it poses to democracy than ever before. There is a vocal lobby that warns about the dangers, but our studies point to a more surprising conclusion. What’s more, our findings can better prepare societies to face the digital demons, telling us exactly what we should be worried about – and what is merely a spectre of our imaginations.
The UK’s 2010 general election was called “the first social media election”. Things have changed rapidly since then. Now, in democracies around the world, it is impossible to imagine parties and politicians campaigning without Facebook ads, TikTok accounts, YouTube videos and WhatsApp channels for supporters. Digital tools have multiple uses. They help raise awareness, bolster pre-existing support and attempt to persuade. But they can also be used to spread misinformation, discourage people from voting and fuel cynicism and discontent.
An important goal of the new science of digital politics is to uncover exactly what political campaigns are doing. Analysing online content is by no means simple: there isn’t a single archive of data and many online spaces are almost impossible to study systematically. However, we do have good access to information on paid-for political advertising – one of the key digital tools linked with microtargeting, misinformation and foreign interference. This shows that online ads are being widely used, with around £7.5 million spent by political parties on Facebook and Google at the 2019 UK general election and a whopping $150 million spent on digital ads by candidates during federal races . But high spending doesn’t necessarily mean campaigners are engaging in digital manipulation. The next challenge is to figure out the impact these new technologies are having on us.
Does microtargeting work?
There is a common perception that digital media tricks voters. This idea has been topical for more than a decade, largely originating from the revelation that a company called Cambridge Analytica was using Facebook data about individuals, without their consent, to sell its services to political campaigns. According to its sales pitch, the firm could understand individuals’ personality types and use them to craft personalised messages to influence their voting. It has never been clear whether Cambridge Analytica could target voters accurately, or the extent to which targeting drove the online adverts it helped deliver. Despite this, it is still widely feared that unknown actors can use personal information available online to control our minds, framing messages that make us dance like unconscious puppets on the strings of our vulnerabilities.
Luckily, there is a rich body of evidence that can help estimate the likely extent and influence of digital manipulation. It indicates that people’s personal data is indeed routinely used by political parties, but microtargeting is surprisingly limited, given the wealth of online information available about our interests, personalities, hobbies, political views, background, demographics, status and social networks. One recent study of 113 national elections in 95 countries found that most campaigns – commonly age, gender or location – when placing adverts. In only a small number of countries did campaigns routinely use four or more targeting criteria. Looking at , our own work has shown that campaigners tend to at certain electorally important constituencies or at men or women in specific age groups, rather than niche audiences.
Hard to influence
Even if microtargeting isn’t as common as we might at first believe, online political adverts could still influence how we vote. Over the past few years, researchers have sought to determine whether this is the case – and they have repeatedly shown that the impact of this type of advertising is small. In one study, conducted over eight months during the 2016 US presidential election, 34,000 people were shown a total of 49 online advertisements. The researchers found that these had an average . In other words, just 7 people in every 1000 said they would change their vote after seeing these online adverts. Another study, published this year, showed that generative artificial intelligence could be used to design adverts tailored to personality types, with the resulting ads being rated as more persuasive when shown to people with the matched personality. However, the – smaller than the variation between different adverts, and a lot smaller than the variation between different people.
Even without political campaigners attempting to manipulate what we see online, the algorithms behind social media feeds shape what individuals see and how they are encouraged to react. Other studies have looked at the impact this might have: for instance, researchers working with Facebook and Instagram examined the effect on users of seeing either the algorithmically determined feed or a chronological timeline during the 2020 US presidential election. The results showed clear differences in the content that users experienced and how they reacted: those with the chronological feed saw more political and untrustworthy content and spent less time on the platforms. Nevertheless, over three months, these differences in levels of polarisation or political knowledge. Again, this indicates that online effects can be smaller than the hype suggests.
Other research has looked at the potential for online advertising to influence voter turnout. One study found that personality-profiled advertisements than generic ones. But real-world studies suggest this effect is small too. One, for example, looked at the impact of a digital advertising campaign designed to persuade people to vote against Donald Trump and for Joe Biden in the run-up to the 2020 US presidential election. Conducted with 2 million voters in five battleground states, the study found that these adverts among Biden leaners by 0.4 percentage points and decreased voting among Trump leaners by even less – just 0.3 percentage points.
It could be argued that because elections are often tight, these small effects might really matter, especially in the extremely marginal seats where we are likely to see the most intense campaign activity. However, it is useful to compare what we know about the effects of digital ads relative to other forms of electoral messaging. Research looking at the impact of door knocking or phone canvassing has shown larger effects, particularly when it comes to persuading someone to turn out to vote. , face-to-face conversation was found to boost turnout by 2.5 per cent, and phone calls from volunteers had a 1.9 per cent effect. Digital ads therefore have a smaller impact than more familiar communication methods.
These results help shed light on the threat posed by digital campaigning. Voters aren’t routinely being targeted in a sophisticated way. When they do see campaign messages, they are hard to influence, and digital political advertising is less persuasive than more established forms of communication. But even small effects could be concerning. So, how worried should we be?
For a start, we need to be sure we are diagnosing the problem correctly. If we view all the research showing that digital campaigning has an influence as evidence of manipulation, it makes the monsters seem scarier than they are (see “Influence versus manipulation”, below). In our view, targeting information at individuals alone isn’t manipulative. Say I am advertising apples and I tell a health-conscious person that apples are a nutritious snack, and I inform an appearance-conscious person that they contain vitamins required for beautiful skin. If they then both buy an apple, it doesn’t mean I tricked either of them into doing something they didn’t choose, nor does it mean I can make anyone buy an apple any time I want.
Defining digital propaganda
But we don’t always see digital persuasion clearly. Each of us holds our opinions, especially on political matters, and these affect our openness to new information. Those with right-wing views are less open to left-wing campaign messages and vice versa. Often people label the tactics and techniques used by their political opponents as manipulative, but are happy with their own party’s behaviour. This inflates the number of accusations of campaigns being manipulative without meaning that more manipulation is necessarily happening.
So, we need to be clear about whether technology is actually having concerning effects or if it only has the potential to do so. The findings we have outlined provide some comfort here – but research in this area also offers lessons. One is that people often view new technologies as more worrying in principle than in reality. Our own research highlights this gap. For example, , we asked 1881 UK voters about every single political advert placed by parties, candidates and non-party campaigners on Facebook during the 2019 general election. Most were deemed broadly acceptable, with just 40 out of 2506 adverts judged to be “definitely unacceptable” by all respondents.
We also know that individuals’ reactions to digital persuasion vary. People tend to be good at interpreting information from known sources, but aren’t necessarily as savvy when it comes to unknown ones. Moreover, that some people are more open to political advertisements from unknown sources, which could make them more susceptible to manipulation.
How to tackle misinformation
There are also aspects of digital democracy that we need to understand better. Existing studies have focused mainly on voting intentions and turnout. We know less about the effect of advertisements on fuelling cynicism and disengagement from democracy in general. More research is required to assess the effectiveness of efforts to spread disenchantment or sow confusion online. We also don’t know enough about . Although studies indicate that , the solutions touted by some catastrophists are simplistic: it can’t just be censored away, designed out or easily inoculated against.
It would be wrong not to worry about the effect of digital tools on democracy, but it would also be wrong to panic. There are undoubtedly some problematic and concerning practices. What’s more, the future is uncertain: just as microtargeting unleashed a new wave of concerns a decade ago, now AI is seen to pose a novel and dangerous threat to democracy. But we must be clearer about the precise problem before we pass judgement on its societal effects.
Besides, digital tools can also be used for good. Democracy is founded on the idea that voters should be informed and that there should be competition between campaigners with differing visions for the future. The use of digital technology to advance these goals is entirely compatible with a healthy democracy. And it does do this. For instance, research published this year found that US voters who agreed to have their Facebook and Instagram accounts deactivated six weeks before the 2020 presidential election , but were no less polarised in their views than those who kept scrolling social media – revealing the value of digital platforms for informing as well as misinforming.
Looking closer at what is happening has convinced us to be less worried about the digital future of democracy. People are harder to trick than we tend to assume. Psychological research indicates that they are also surprisingly reasonable and rational. You may not like the choices your fellow citizens make, but they have their reasons. The real worry isn’t digital propaganda, but that we stop believing in our compatriots and give up on democratic persuasion altogether.
Influence versus manipulation
Philosophers have spilled a lot of ink arguing about what defines manipulation. A core feature is deception – that the persuader hides either their true motive or something else about the persuasive message. This should immediately tell us that manipulation is a matter of degree, because even the most honest person can't tell you everything about their intentions or what they know and when. And even the most honest person will sometimes try to persuade you of something. That doesn't mean they are manipulating you.
A simple example illustrates this. Imagine you offer a crowd of 100 people an apple. You would probably get some takers. But that isn't manipulation: it is likely that the people you persuaded already liked apples or were getting peckish or both. It would be misleading to claim this as evidence that you were able to control people's minds. Likewise, an experiment finding that online ads affect people's vote choice doesn't necessarily prove that digital advertisers manipulate voters.
Kate Dommett is a professor of digital politics and Tom Stafford is a professor of cognitive science. Both are at the University of Sheffield, UK
