Since late October, Facebook has come under scrutiny for changing its policies surrounding political ads. Previously, to run a political ad on Facebook, the ad had to be factually correct. Facebook would review political ads before they went live, and on an ongoing basis once they’re put live, removing any that it found to contain misinformation.
Facebook repealed that policy in late October, arguing that it doesn’t want to censor political speech. Zuckerberg argued for the policy change by explaining that as a private company, Facebook wants to moderate content as little as possible.
Facebook’s argument also hinges on the fact that other forms of ads, such as those that appear in TV and print, aren’t subject to fact-checking. If these aren’t subject to fact-checking, the argument goes, then why should Facebook have to fact check ads that are run on its platform?
Facebook’s position on political fact-checking has come under criticism for a number of different reasons. The main argument against it is, unsurprisingly, that it allows disinformation to spread freely over the platform.
If Facebook doesn’t assume any sort of responsibility for fact-checking this content, then people can come to believe things which are factually incorrect. With elections approaching in both the US and UK, two countries where social media plays a large part in informing voter beliefs, there’s a danger that misinformation spread via Facebook ads could affect the results of these elections.
This worry is particularly prevalent in the age of micro-targeting. When political advertisers can precisely target specific groups of people, there’s a heightened danger of factually incorrect ads being used to radicalize people’s political beliefs.
Micro-targeting played a significant role in the 2016 election. In the run up to the election, the Trump campaign used a array of voter profiles to group Facebook users, and serve them ads designed to resonate with them specifically. Given how effective this was believed to be in 2016, it’s only going to become a more effective strategy in 2020 with Facebook’s refusal to fact-check content.
In late October, Twitter responded to Facebook’s change in policy by moving in the opposite direction. In the hours before Facebook’s earnings call, Twitter CEO Jack Dorsey announced that the platform would be banning all political advertising.
It was a savvy move from Twitter, not least because only a very small percentage of Twitter’s ad revenue comes from political advertising. During the 2018 US midterms, the platform generated just $3 million worth of political ad revenue, about 0.1% of its total ads revenue for that year. By banning political ads, Twitter has lost an unnoticeable amount of revenue.
For that small drop in revenue, Twitter is likely to make gains in consumer trust. Coming just a month after Twitter revealed that it had been using recovery emails and phone numbers to target advertisers, Twitter is likely to gain back some (if not more) of the trust that was lost following that story.
By trading a minor amount of revenue for consumer trust and the chance to land a blow on one of its biggest competitors, Twitter’s choice to ban political ads was a well-made business decision. As commendable as this decision was from a business perspective though, Twitter has come under fire for it from a number of different fronts.
One of the main arguments against Twitter’s move to block political ads, and in favour of political ads in general, is that political ads can be democratising. By offering an avenue in which politicians can reach voters, political ads allow voters to make more informed choices. This point is particularly relevant for digital political ads, due to the fact they’re self-serve.
Self-serve media is any form of media which someone can go and distribute themselves. Digital ads are typically self-serve in that you don’t need permission from a 3rd party in order to distribute the ads; anyone can open up a Facebook Ads account and start serving ads.
This is in contrast to traditional media channels like TV, radio, and print. The barriers for entry are typically much higher on these channels, and that’s just for paid content. If you want to reach an audience organically through a channel like TV, perhaps through a feature on a news channel, you have to already be a well-established politician.
Traditional media channels in this sense favour the incumbent political class. The high barrier to entry makes it difficult for non-establishment politicians to use these channels to their full effect. Digital ads, on the other hand, offer a much more level playing field. The fact that anyone can get started on a channel like this means that (finances aside) political challengers have just as much chance of being heard as established politicians.
By banning political ads, the argument goes, Twitter has shut down an important channel of communication for non-establishment politicians. Their decision favours incumbent politicians, who are likely to receive greater attention, (and therefore reach) from traditional media channels.
This has been the main argument against Twitter and its decision to block political ads from its platform. There is another angle from which the company has been criticized though, one which claims that the boundary between political and non-political ads is far too blurred to draw a meaningful distinction.
Take issues like abortion. If an organization, either in favour of or against abortion, decides to run ads advocating their position, is this political? In some sense it is; these are ads which attempt to swing public opinion on a heavily politicised topic. In another sense though, these ads don’t seem to be overtly political. They may not reference, support, or attack any political party.
Should these ads be considered political? If we consider them political, then we risk infringing on freedom of speech; shutting down a group of people’s right to share their opinion on a topic just because it’s politicised. If we don’t consider these ads political though, then we open up an avenue for politicians to skirt restrictions on explicitly political ads.
If politicians aren’t allowed to run ads in support of themselves, they could still be allowed to run ads advocating for positions which they also support. Running ads in support of these positions would be of benefit to said politicians, particularly if they’re the most well established politician advocating for that position. If this happens, voters may favour the politicians who are advocating for these positions.
And so Twitter finds itself in a dangerous dilemma. Does it allow ads on political topics, thus allowing political parties to serve ads attempting to change voters’ views on topics relevant to that party? Or does it blanket-ban anything remotely political, and in doing so risk accusations of heavy handed censorship?
Whatever road Twitter goes down, it seems that there are always going to be difficulties with trying to enforce an outright ban on political advertising. But Facebook’s approach of letting political ads run freely, without regard for their accuracy, is no better.
As we saw earlier, Facebook’s policies allow politicians and their parties to run micro-targeted misinformation campaigns. In electoral systems like those of the US and UK, where a couple thousand votes in a few states or constituencies can decide an election, these campaigns have the potential to sway an election.
Whether we ban political ads or let them serve without restriction, there are problems at both ends of the spectrum. A potential solution which has been gaining ground is to allow political ads to serve on social ad platforms, but to restrict the ways in which they can be targeted. In particular, this line of thought suggests that political ads should be allowed to run as long as they can’t be micro-targeted. This means no granular voter profiling and no behavioural targeting.
The benefit of this approach is it means that political advertisers won’t be able to use the power of Facebook and Twitter’s ad platform to reach precise segments of people. Not being able to target users so precisely means that political advertisers will be less effective at spreading disinformation amongst those most likely to believe it.
A secondary benefit is that it will be more difficult for political advertisers to saturate certain audiences with their content. Saturation refers to when an advertiser shows a particular ad to an audience over and over again. Saturation is typically something to be avoided in commercial advertising, as it means you’re wasting ad spend on people who’ve likely already heard of your product.
In political advertising, saturation could potentially be more desirable. This comes down to something called the illusory truth effect, a well studied psychological phenomenon which describes how people are more likely to judge something to be true if they’re familiar with it.
One of the most relevant studies on this was conducted in 2012 by Central Washington University. In this study, participants were exposed to fake news stories which were presented as factually correct. Five weeks later, participants who had been exposed to the fake stories were more likely, compare to those not exposed, to judge the stories as true.
Additionally, many of those exposed to the fake stories were not able to recall hearing of the stories as part of the study, suggesting that we can easily mis-attribute the source of false beliefs.
What this tells us is that political advertisers can exploit the illusory truth effect by serving false information repeatedly to users. The tendency that users have to mis-attribute the source of their beliefs means that many will forget they had originally come across this information via a paid ad. This is important as paid ads may be a less credible source of truth, and so by forgetting that the source of their beliefs was a paid ad, users may be more likely to believe this disinformation.
Micro-targeting allows political advertisers to ensure specific users are repeatedly exposed to specific messaging, which can then have a significant effect on their beliefs. By banning political advertisers from using micro-targeted campaigns, the argument goes, we can reduce the risk of these advertisers saturating audiences with their messaging, and getting voters to believe falsehoods.
I fully expect that Facebook will ban micro-targeting in the near future. It’s already banned micro-targeting for any ads to do with credit, employment, or housing opportunities, following a lawsuit at the start of 2019. As a result of this lawsuit, Facebook has required all advertisers to label relevant ads. The advertisers are then unable to use any granular targeting methods for these ads, and instead must target very generic audiences which often contain hundreds of millions of users worldwide.
Facebook already has all the infrastructure in place to ban micro-targeting for political ads. All it would need to do is include political ads into the list of categories which it disallows micro-targeting for. As pressure mounts on Facebook in response to it’s recent changes in ad policy, I expect Facebook to eventually acquiesce, and disallow micro-targeting for political advertisers.
There is a serious issue though in thinking that removing micro-targeting alone will stop political advertisers from running effective disinformation campaigns. To understand this issue though, we need to understand a distinction between two different types of advertising.
Traditional advertising, be it advertising through TV, billboards, radio et cetera, is purchased through what’s termed a reach buy. You pay a certain amount in order to reach a certain amount of people. For each of these mediums, the number of people exposed to them is fairly predictable. TV networks know how many people are likely to watch a show, billboard owners know how many people walk past their billboards each week, and so they’re able to charge a fixed amount to advertisers who want to use these mediums.
What’s important about reach buys is that the advertiser doesn’t get any real control over who sees their ads within the mediums that they advertise on. Sure, if they want to reach a younger audience for example, they can advertise on a TV show that young people watch, but the advertiser can’t get more granular than that. They have to target broad audiences rather than individual people.
This is in contrast to how digital advertising is typically bought. Digital ads, whether they’re on Facebook, Google or wherever else, are typically bought through an auction. To understand how this works, let’s say you make a Google search for the term flights to paris and some ads appear at the top of the results page. What Google has done while you were waiting for your results page to load is to hold an auction between every advertiser in the world who wants to show you an ad on that search term.
Different advertisers will bid differently on your search based on a whole range of factors. These could be your age and gender, which sites you’ve been on previously, or whether you’re about to get married and are therefore possibly looking at honeymoon destinations. In this way, advertisers can optimise who their ads are shown to, and decide who they actually want to spend money reaching.
Auction-based advertising is arguably even more powerful on Facebook, where advertisers can optimise their ads to be shown to users who are likely to respond in a certain way. One way of doing this is to optimise your ads to be shown to users who are most likely to engage with them, either by clicking on them or reacting to them. This tactic was almost certainly employed by the Trump campaign in 2016.
In allowing advertisers advertisers to only pay for the users they want to reach, auction-based advertising effectively offers a way around any ban on micro-targeting. Political advertisers would have to start with a larger audience to target, but once they’ve learned what sort of users are engaging with their ads they can hone their targeting to just focus on those users.
If politicians are running incendiary disinformation campaigns, there will be a certain sort of user who is likely to engage with these ads. By optimising for users who are likely to engage with these ads, political advertisers can skirt the limitations of not being able to micro-target. Banning micro-targeting will have only a very limited effect on how effective these campaigns are.
In order to really restrict political advertisers’ ability to run these sorts of campaigns, Facebook and other platforms need to restrict how these advertisers can optimise their campaigns. Platforms need to remove options for advertisers to show their ads only to people likely to engage with them.
These platforms already have options in place for this, which allow advertisers to reach people without regard for who they are or how they’re likely to respond to the particular ads. By forcing political advertisers to use these options, these advertisers will find it exponentially more difficult to run granularly targeted disinformation campaigns.
If platforms make this change, the digital political ads landscape will start to look much more like it does on the traditional media channels, such as TV and radio. The inability of traditional media channels to micro-target, or optimise ad serving to particular users, means that advertisers have to craft a message that appeals to all audiences. Disinformation ads are likely to be less effective in this environment, not just because they’ll be heard by a broader audience which is more likely to call them out, but also because they can’t saturate specific users’ media consumption in the same way; it’s harder and more expensive to saturate a big audience than it is a small one.
One of Zuckerberg’s defences for Facebook’s refusal to fact-check ads was that if traditional channels don’t have to, why should Facebook? This argument doesn’t work precisely because Facebook is so different to traditional media channels. But if Facebook implements the suggestions above, of banning micro-targeting and ad optimisation methods, it will become a channel much more akin to traditional media channels. Only then will Facebook be able to hold itself to the same ad-checking standards as these traditional media channels.