A few days ago, I found myself talking to a woman I had just met, running through a list of facts about her family.
The week before, Demos had been challenged by the BBC to find out as much as possible about a volunteer, using only public, online data, and starting solely from their email address and a protected Twitter handle. We weren’t allowed to spend any money, or put more than a few hours into the research. “She doesn’t think there’s much out there”, we were told. “See if you can surprise her.”
It was her mother’s history, gleaned from a public social media account, which first raised an eyebrow from our volunteer. We were able to guess, correctly, which MP she had voted for in the last three elections, and the way in which she had cast her referendum ballot. We walked through the causes she’d supported over the years, and the campaigns she felt most strongly about. The internet knew her current address, the house she had moved there from, and how much the latter had sold for. It knew the shape of her signature, and her husband’s.
This experiment does not replicate the ways in which political parties prepare for a campaign. Even the best funded groups are unlikely to have the resources to conduct their own, detailed searches on every voter. The tools for doing similar work at scale, however, already exist, and are being used in political campaigns.
Earlier this year, Demos was commissioned by the ICO to produce a report into the future of political campaigning – looking at the ways in which political groups currently use data, and how this might change given advances in artificial intelligence, big data and targeted advertising. The report, freely available here, found that by the time of the next general election, emerging technology is likely to have changed the face of the UK’s campaigns.
Many changes are already underway. During the EU referendum, Vote Leave are reckoned to have run around a billion targeted adverts, purchasing access to the swathes of data held on citizens by Facebook and companies like them in order to better reach voters. UK Political parties maintain their own databases – in tools like Nationbuilder and NGP VAN – which store any collected information on wards, streets and people which could be useful in a campaign. Often this data is connected to information purchased from private data brokers; Adobe, Oracle, IBM.
In the near future, campaigns will be able to draw data from an ever more inventive range of sources. One American firm touts the ability to target the attendees of a given political rally or march by geolocating phones present at the event, and allowing campaigns to send adverts to people’s computers when they get home. In the next few years, advances in the ability to track sentiment are likely to make it possible to detect emotion from tone of voice or expression, opening the door to messages delivered when voters are detected to be the most emotionally susceptible. In many cases, inferences could be made from seemingly innocuous data – one patent taken out by General Motors describes a method by which ‘vehicle trace data’, such as the height of your seat or the radio station you listen to in the morning, could be used to infer your likely age, gender, and income.
Many of these capabilities will be genuinely helpful to political parties in planning and running campaigns. If carefully used, there is an argument that developing a clearer picture of the electorate could lead to more considerate policy-making. Accurate targeting can also help campaigns better direct scarce resources, and the fact that these technologies are often relatively cheap can even the playing field towards smaller, less well-funded campaigns.
There is a danger, however, that these advances will lead to the gradual atomisation of the electorate; each voter addressed individually by parties, playing on their particular makeup of list of causes, habits and emotion. In this atmosphere, campaigning becomes not only more difficult to regulate, but also more difficult to discuss – it is hard to generate a shared idea of what a party stands for if your neighbours and friends are exposed to completely different messaging during the run up to an election.
We may, of course, decide that all of this is worth it – that we don’t mind helping the the causes we care about by donating our data, and that we trust regulators to prevent serious abuses. But consent, and effective regulation, requires knowledge. Both we as citizens and the organisations designed to protect us must be aware of the data which is out there about us, and how it is being used in campaigns. We must be able to stay on top of new developments, often occurring in response to regulation. If the raised eyebrows of our volunteer last week are an indication, we still have a long way to go.
Read our report The Future of Political Campaigning.
 C Cadwalladr (2017), ‘British Courts may unlock secrets of how Trump
campaign profile US voters’, The Guardian, Available from:
 G Windsor and S Murphy (2014), ‘Big data and the 2015 UK General Election: digital democracy or digitally divisive?’, Available from: https://www.nesta.org.uk
 J Chester and K Montgomery (2017), ‘The role of digital marketing in political campaigns’, Internet Policy Review
 K Caswell, ‘Three Political Marketing Strategies: Using El Toro for Digital Advertising Political Marketing Strategies’, Available from: https://www.eltoro.com/politicalmarketing-strategies/
 Breakthrough Analysis (2016), ‘Text, sentiment & social analytics in the year ahead: 10 trends’, Available from: https://breakthroughanalysis.com
Rachuri et al (2010), ‘EmotionSense: a mobile phones based adaptive platform for experimental social psychology research’, UBIComp
 GM Global Technology Operations LLC, (2018). Learning driver demographics from vehicle trace data. US9928524B2.