This is the 2nd year that of the Award for Best Use of Data Science to Achieve Social Impact powered by Deutsche Bank. For 2018 we welcome submissions from individuals and organisations across Europe, this category is for data science projects which have had a strong positive impact on social justice or environmental issues affecting their local community or the broader world around them. Topics included projects which seek tackle issues linked to the environment, poverty, social and economic disadvantage. 

One of this year’s finalists is the Transparent Referendum Initiative, which was set up to prior to the recent Referendum in Ireland on reproductive rights, with the aim of making it open, honest and fair, by building a database of all social media ads targeting voters and making them open to scrutiny and fact-checking.

We caught up with Liz Carolan founder of TRI to tell us her story, best of luck to Liz & the team on September 7th!

This year has felt like a whirlwind. Back in December, I sketched out an idea in a blog post. Within six months there was a community of over 600 volunteers feeding investigative journalism and political action that led to a major climbdown by two tech giants and commitments to reform by a national Government. The process was a steep learning curve, and one I am currently working on extracting lessons from for building data driven projects that deliver social impact.

The original idea for the project was an outline of a problem and a seemingly straightforward solution.

The problem: Ireland was due to hold a referendum on abortion in May which was already gathering a lot of international attention. In the US and UK, investigations were uncovering that social media platforms had allowed both disinformation (“fake news”) and illicit financial flows (overseas or otherwise unaccounted for funding) to influence voting. Those investigations showed that online ads can target information – videos, posts, articles etc. – at small groups of people, based on sophisticated data-driven analysis. But the ads are only seen by those groups, and then vanish. This left no way for journalists, campaigners or official bodies to know what information was circulating, who was behind it, and where the financing came from. Months, even years later, we are still working to try unravel what happened in these two countries.

The solution: build a real-time public database of political ads that were being shown to voters on social media sites like Facebook, so that they could be exposed to the same scrutiny, fact checking and source tracing as other political messaging and spending.

Data, and the algorithms that use it, formed part of the problem, but perhaps it could also form part of the solution.

The first and most important step was to work out what was needed to be able to deliver the project. We started by building a core volunteer team with an understanding of digital advertising (Peter Tanham), social media campaigning (Craig Dwyer) and data transparency and governance (me). But with no funding, a project like this is really only a platform for pooling the volunteer effort and good will of a wide ranging group of people.

In that regard, we were overwhelmed by the willingness of people to contribute. It helped that the conditions were right – we had a defined timeline, a pressing social issue, and concrete asks for people. We also made sure that our branding and messaging were consistently professional, and that we were open about our challenges and constantly solicited feedback.

The first key partnerships we built was with WhoTargetsMe, a tech non-profit who had already developed a plugin that could scrape ad data when it was installed as a Chrome or Firefox extension. This gave us a technical solution to the issue of collecting Facebook ads. We put out a call for volunteers to download the plugin, and over 600 people did. Many of these also sent us screenshots of other advertising they were across other platforms, such as Google and YouTube.

We began by using very unsophisticated methods of filtering and cleaning and publishing the data, but we were able to build a partnership UCD’s Dynamics Lab, and data scientist Killian McLoughlin came on board as part of his MSc internship. This both freed up our time to focus on getting the data used, and helped to build trust that the data was being managed properly and independently.

With the data in the public domain (we would publish metadata on over 1600 ads by the end of the campaign), we saw a clear path to impact – journalists would analyse it, identify wrong doing, expose it, and there would be repercussions. Having built individual relationships with journalists, we were eventually able to build a collaborative network of journalists, together with experts from Storyful and elsewhere, who were actively pursuing leads and providing scrutiny of the campaign. (One member of that group wrote about the experience for Foreign Policy).

This led to stories in The Times, CNN, Buzzfeed, BBC News, The Guardian, The New York Times and a highly impactful report on Channel 4 News, as well as countless other outlets. The pressure of this scrutiny, combined with efforts by members of both houses of Ireland’s parliament, led to both Facebook and Google limiting advertising during the referendum campaign. In the months since the referendum, as we have continued to engage with Irish authorities, it has also led to commitments to reform Ireland’s regulations so that online political ads can no longer allow disinformation and illicit financial flows to influence our referendums and elections.

No doubt our project was helped by timing and luck – the Cambridge Analytica scandal broke right in the middle of the Irish campaign, adding massively to pressure on the tech giants. But even prior to that escalation, I was astounded by the willingness of people to join forces and give their time and energy to doing something positive and constructive for democracy. A key insight I will take with me to my future work is that if you give people with an opportunity to contribute in a meaningful way to an impact focused project, they will jump at it.

Since the referendum we have been approached by activists around the world interested in doing similar projects, and are supporting colleagues in countries like Sweden to replicate what we did in Ireland. The team are continuing to work with Irish officials and parties to ensure regulations in Ireland are updated to cover online advertising. We believe that here Ireland has an opportunity to lead the way in figuring out how this can work. The world is watching

The DatSci Awards will be Celebrating Data Science Talent in a unique Awards Ceremony that is interactive & fun!  Anyone in the Data Science Community is welcome to join and it is a great opportunity for you to connect with some of the finest minds and organisations in the industry. Purchase your ticket here.

(L-R Craig Dwyer, Liz Carolan, Killian McLoughlin)