On 25 May, Ireland voted by a two-to-one margin to allow its parliament (the Oireachtas) to change the constitution in a way that would legalise abortion. In this post, Liz Carolan discusses the role of digital media in the referendum campaign, the challenges it poses for democracy, and potential solutions to the problems she observed.

Background

When it was announced that there was to be a referendum on abortion in Ireland, not many people anticipated a landslide; I certainly did not. I had spent the previous five months trying to monitor the financial and information flows behind digital political advertising, witnessing attempts at overseas interference, disinformation campaigns, and unregistered spending. With a tight result predicted by both polls and campaigns, I thought it was naive to think that the digital campaign would not be a deciding factor.

The Irish vote had real and concrete implications at home, but there were international stakes as well. Ireland’s constitutional ban on abortion was held up by activists in the US and elsewhere as an example to be emulated. Pro-choice advocates globally were watching too, as Ireland’s vote to legalise same sex marriage in 2015 had emboldened equal marriage activists around the world.

So long before polling day, threats to the proper functioning of the referendum were evident to those of us spending time thinking about technology and democracy. The first threat was that overseas or untraceable financing would be used to try to influence the vote. The second was that deliberate disinformation campaigns could spread untruths, disparage campaigners, and polarise or isolate voters. The last was that a large amount of campaign spending could happen under the radar.

Digital advertisements are particularly interesting because they bring together money, information, and the algorithms that determine who sees what, and importantly who doesn’t. They are often only seen by those targeted with them and they are ephemeral, with the ability to appear and vanish without leaving a trace. They had been an avenue for the alleged overseas interference and deliberate disinformation campaigns during the Brexit referendum and the 2016 US presidential election.

The work of the Transparent Referendum Initiative

Looking at the ongoing investigations into these cases, it appeared that investigators and legislators alike were having challenges even knowing what had happening online during the voting period. So in February some colleagues and I launched the Transparent Referendum Initiative (TRI). We decided to build a database of as many ads as we could, to make it available to as many people as possible, in as close to real-time as possible. We did this so the ads could be exposed to scrutiny, fact-checking, and source-tracing, so that any media or regulatory response could be swift and contemporaneous, rather than retrospective.

We have no funding for our work, so we used openly available resources to run the project. We partnered with the UK organisation Who Targets Me (WTM), which had developed a tech tool that captures ads shown to people who install a plugin in their web browser. 600 Irish voters signed up to use WTM and anonymously pool every ad they were shown on their desktop or laptop, so that we could pull out and build a database of ads shown to at least a segment of the Irish population. We also encouraged people to send us screengrabs of ads they were being shown on social media and across other digital platforms.

The evidence collected by TRI

We were sent scores of screenshots of ads on YouTube, and across the web served by Google and other ad services. Through WTM we logged 1535 distinct Facebook posts from mid-February up until polling day on 25 May. We shared the Facebook ads publicly in an open online database that we updated every few days. Ads we were sent as screenshots were logged and passed on to journalists who could verify and investigate them.

In this partial snapshot we identified a lot of activity that we believe violated the principles of our electoral rules. Ireland has some quite strict rules when it comes to campaigning, upholding the idea that political spending should not be allowed to be limitless, untraceable, or from outside the country. TV and radio ads are banned, and all broadcast media must present both sides of a debate equally. Groups must register if they are receiving donations to campaign, and they are banned from accepting donations that are large or that originate overseas. Posters and print ads must contain information about who has paid for and printed them. However, it quickly became apparent that no institution had the capability to respond to any of the digital activity we were seeing.

The ads we captured with WTM came from 317 different groups, two thirds of which were not registered as campaigners or even loosely affiliated with registered groups.

17 different groups paying for advertising to reach Irish voters were registered in other countries, including the US, Canada, and various parts of Europe.

A further 26 groups found using WTM were untraceable – the Facebook pages pushing the ads contained no information about the location, identity, or intent of the person paying for them. Many of the ads sent to us via screenshots also fell into this category.

Amongst these ads we saw videos, links and other content attempting to mislead voters, discredit political figures and groups, mimic official or neutral information sources, gather voter data, and share disturbing images.

The role of regulators, the media, and politicians

One particularly troublesome ad that was reaching large numbers of voters at the height of the campaign exemplifies the inability of our regulatory institutions to respond. It was shared by an untraceable Facebook page and linked to a US-hosted blog, spreading untruths about the funding for one of the campaigns. It was timed to coincide with a fundraising drive by the campaign they were working to discredit.

We spotted the ad, and sent it to five separate entities – SIPO (the regulator of political finance), the advertising standards body, our broadcasting authority (which maintains balance in broadcast and upholds the ban on TV and radio ads), and the Referendum Commission, a temporary body set up to administer the vote. Each replied saying they had no authority to act. Facebook also told us they don’t fact-check content.

With no regulatory response, media scrutiny proved key to ensuring there was some measure of oversight of the process. Many news organisations investigated the leads in our data, publishing stories informing voters of what was going on before they voted, including the Irish Times, Ireland’s Sunday Business Post, The Guardian, Channel 4, the New York Times, CNN, and BuzzFeed. The Times (Ireland Edition) ran multiple stories uncovering misinformation by ‘No’ groups, including spreading bad science about links between abortion and cancer. Wired reported on an anonymous group spreading racist footage taken out of context.

Members of both houses of the Irish parliament began calling for action, and criticising the companies. TRI wrote to key ministers urging them to ask the companies to release information about who was paying for ads, and how much was being paid. With almost all of the investigative journalism showing poor behaviour on the ‘No’ or pro-life side, but not on the ‘Yes’ side,  it was clear that no state institution felt able to act without appearing to interfere in the referendum process.

The role of Facebook and Google

As pressure mounted, Facebook and Google took the unprecedented step of limiting or withdrawing their advertising services for the referendum a little over two weeks before polling day. Facebook announced on 8 May that it would prevent overseas groups from buying political ads targeting Irish people. The next day Google went one step further, imposing a ban on any group buying political ads across its platforms, regardless of their country of origin. This appears to be the first time these companies have done anything like this.

The reaction by ‘No’ campaigners, in particular to the Google decision, was swift and angry. They declared that by removing their ability to implement part of their digital strategy, Google was engaging in election interference. They claimed that the result was ‘rigged’. This reaction was amplified by American conservative media. The National Review carried the headline ‘Silicon Valley Deletes the Pro-Life Campaign in Ireland’. The bans represented to them further evidence of political bias in these too-powerful private companies, and efforts to suppress conservative speech.

Neither Google or Facebook fully explained why they took their respective decisions, citing general concerns about ‘electoral integrity’. With polls tightening, the companies may have been motivated by fear they would be blamed if the referendum failed to pass. A narrative that they had helped a ‘No’ vote by allowing overseas involvement would have been very bad for their public image, already damaged by the ongoing investigations into interference in the 2016 US election.

Either way, these unprecedented acts of self-regulation appear to come from a conclusion that the absence of regulation poses reputational and therefore business risks. Each company’s policies state that they will comply with the laws of the countries within which they operate. But recent scandals involving online campaigning have shown that companies are expected – by the public, their own staff and shareholders – to do more than just comply with the law. They are realising that operating in a ‘Wild West’ can also mean mob justice. 

While laws take years to write and pass, code can be rewritten, or rewrite itself, within hours, minutes, or seconds. It is becoming clear that the ways our electoral management systems are set up are potentially not fit-for-purpose for the digital age. However even when these companies attempted to act to restrict the use of their services, they failed. In the days before the vote, The Guardian investigated ‘No’ ads appearing on their site against their knowledge. Neither they nor their advertising partner could work out how they got there. Google advertising infrastructure was used to serve these ads. Facebook also found it difficult to prevent overseas groups from advertising. Relying on artificial intelligence, the technology that Mark Zuckerberg referred to repeatedly in his parliamentary hearings in Brussels and Washington as a way to deal with these issues, did not work. Despite a genuine commitment from the Facebook team in Dublin, some overseas ads got through while some Irish groups were blocked.

What should be done to deal with the problems of digital campaigning?

Focus on this debate momentarily faded into the background when the results started to come in showing what that Ireland had voted for reform by an unanticipated two-to-one margin. However with further referendums planned for later this year, as well as the potential for both presidential and parliamentary elections in the near future, there is increasing pressure on the government to take action now.

We, along with many political commentators and representatives, are calling for a minimum threshold of transparency for all online political advertising, in real time, backed by legislation. Providers of advertising spaces should provide an open application programming interface (API) of the identity, spend, targeting, and content of any paid advertising in the run-up to an election or referendum. The institutions who oversee elections and referendums will then need to be empowered to respond as technology adapts. Facebook and YouTube ads are the problem of today, Whatsapp and other closed platforms are a growing issue we will be hearing more of in the coming months and years. There are also many challenges we have not anticipated yet. However if we want to work out how to address digital threats to democracy, we need to take a look at some of the broader structures and institutions that make our democratic processes vulnerable to them.

In Ireland, in the end, the campaign itself, including the digital campaign, may have actually had little impact. Analysis of the RTE exit poll found that just 12% of respondents said they made up their minds during the referendum period, with 75% saying they ‘always knew’ how they would vote. While this is likely an exaggeration, it does show that if there were broader factors in the overall running of the referendum, that may have made the process less exposed to online manipulation.

Firstly, the referendum itself was the end of a long, deliberative process that included a citizens assembly and high profile parliamentary committee, where the evidence and viewpoints that would dominate the campaign were debated and discussed at length and in a respectful forum. Through these processes, key individuals – included elected reps – publicly changed their minds. Secondly, the government took a risk in publishing an outline of the bill they would put forward in the case of a ‘Yes’ vote, providing greater clarity on what people were voting for.  And thirdly, as political commentators and representatives repeatedly told me, in Ireland the ground campaign still rules. Ireland’s small, multi-member district voting system (which combines proportional representation with the Single Transferable Vote method) means that voters are used to – and in some cases expect – candidates or campaigns to present their position in person on their doorstep. Anecdotally, friends who were campaigning told me that when knocking on doors during the campaign, voters regularly said they had been waiting for them as they had a list of questions they needed answering.

None of these factors, however, leave any room for complacency when it comes to ensuring there is a proper regulatory framework, backed by legislation, in place to cover online advertising. Here Ireland has an opportunity to set the bar high when it comes to finding ways to making sure that digital campaigning is open, fair and honest. The world is watching. If the Irish government do this well, others will surely follow.

About the author

Liz Carolan is an adviser on transparency and governance who has supported reform efforts in over 30 countries. She is co-founder of the Transparent Referendum Initiative, a project advocating for improved regulation of political campaigning in Ireland.