“Elections are becoming increasingly ‘datafied’.” This is according to the three authors of ‘The Future of Political Campaigning,' a report commissioned by the ICO and conducted by the cross-party thinktank Demos. Following a detailed analysis and examination of innovative uses of technologies by marketers and advertisers, the authors identify seven trends in the use of data and analytics technologies by political campaigners with examples of how they are being used now and predictions of how they could be used in the future.
The first trend it calls out is audience segmentation, which allows audiences to be divided into smaller groups on the basis of granular information about their demographics, behaviour and attitudes. Through this, campaigners will be able to get the most persuasive messages in front of undecided voters or those in swing constituencies. The authors predict that data from IoT devices could eventually be twinned with machine learning to give an even more detail picture of potential voters. Influencers may be the next voter demographic that political campaigners target, given their ability to "advertise" their political opinions to their friends, it predicts.
"The direction of travel is towards a customer segment of one."
The second trend is cross-device targeting, advertising to the same person through many different platforms. The Democratic National Committee worked with data services firm Experian and political data company TargetSmart Communications to aim video ads, mobile and desktop display ads at specific voters. “The direction of travel is towards customer segments of one,” said the authors. The report states that, while they are not aware of political campaigners using IoT advertising, it is plausible that they will start looking at such techniques.
Trend number three is the growth in the use of psychographic techniques that aim to determine the specific personality types, attitudes, values and interests of users. Political campaigners could then use that personality data to produce content tailored to that person. Experian Marketing Services is already offering psychographic techniques it notes. However, the authors point out that there is little evidence as to its effectiveness in influencing political choices. If in the future this kind of technology is introduced to the High Street, political parties might be able to analyse the facial expressions of people watching television adverts or political debates and tweak the messages they are sending out according to the reaction of the viewer.
The fourth trend is the use of AI to target, measure and improve campaigns. AI could outperform human strategists in working out who should be targeted and when and with which content in order to maximise the potential to persuade. The authors said that these technologies could be used to monitor and improve the performance of political campaigns through the A/B testing of adverts which iteratively improves and targets the message more effectively.
"Technology ... that hallucinates faces has the potential to promote misinformation."
The fifth trend -and perhaps the most worrying - is the automatic generation of content, which can be used benevolently or in a malign way. The authors state that Natural Language Generation tools could be used alongside algorithmic targeting to generate content automatically for unique users based on insight around their interests and concerns.
“A system could use trending topics, personal data and an understanding of the interaction between these to generate bespoke and nuanced advertising content,” say the authors. This is currently being used by the non-partisan chatbot Hello Vote, which voters could use to check whether they registered and help them get the required ID.
However, AI-generated content has been used to create misinformation and disinformation and social media bot accounts have flooded social platforms with false information. And it could go further to fuel fake scandals. “Technology which generates photo-realistic images, imitates real voices and hallucinates faces has the clear potential to promote misinformation, providing false ‘proof’ that politicians have said or done something scandalous,” th eauthors warn.
The sixth trend is the use of personal data to predict election results. A 2017 study found that liking a politicians’ public Facebook posts can be used as an accurate measure for predicting voter intention. The authors said they anticipate increasing use of social media sentiment analysis to both gauge reception to a candidate’s speeches or events; and to predict the election outcomes.
The seventh trend is delivery via new platforms. Through the growth of digital video, wearable tech and virtual reality, VR campaigns could be used to open political debates. IoT devices, particularly wearables, are seen as a growth area for marketing especially when combined with location data and could be used by political campaigners in the future.
The authors then went on to set five key challenges that these new technologies pose when applied in political campaigning.
"Campaigns will be incentivised to hold or obtain more personal data."
Privacy is the first principle as the authors note that, “campaigns will be incentivised to hold or obtain more personal data on individuals, and to collect as much diverse data as possible in order to maximise the effectiveness of their messaging.”
User consent and knowledge is another issue for concern, as consumers may not knowingly consent to their data being analysed if they do not even know it is being held or by whom. ”The more complicated and automated the process of data use and targeting becomes, the more difficult it will become for users to ask for a clear explanation about how their data is used, and to know whether they can ask for it back,” wrote the authors. They went on to say that the principle of informed consent will be increasingly difficult to apply considering the complexity of big data and algorithmic technologies, and the difficulty of scrutinising AI processes.
The use of Natural Language Generation to automatically create tailored content for each voter is another concern. “It could result in inappropriate, inaccurate or prejudicial adverts appearing,” the report cautions. The existence of this type of advertising could undermine public confidence in the ability of regulators to ensure political campaigns are run fairly and increase public distrust of political parties.
"There is a risk relating to the transparency and political accountability of campaigns."
Regulators are advised to consider how high volumes of messages might be stored and shared in a way that they can be checked. “Unless this can be achieved, there is a risk of growing concern relating to the transparency and political accountability of campaigns."
Psychographic techniques are likely to improve with the creation of very large and cross-referenced datasets, bringing with it the possibility of emotional manipulation. Linking personality survey data with data sets from IoT devices, location histories and social media will help campaigns build up correlations between personality types, moods, psychological states and patterns of behaviour. This could be used to target potential voters at the time and place that they are most receptive or perhaps most susceptible or vulnerable.
These trends and challenges are followed an in-depth analysis and examination of the ways in which data and analytics is being used by advertisers and marketers. The authors reference the Donald Trump US election campaign in 2016 which made the most of big data analytics technologies, as did the Vote Leave campaign in the EU referendum and the Labour Party, which used an in-house tool that combined voter data with Facebook information to send locally-relevant messages.
They state: “Advertising and marketing techniques are being offered by a network of private contractors and data analysts, offering cutting-edge methods for audience segmentation and targeting to political parties all over the world.” Some of these companies and the services they offer are mentioned in the report, such as WPP’s Xaxis Politics. The report made many references to reports by academic researchers, including Moore and Tambini, whose research indicates that micro-segmentation could inadvertently result in consumers being sent messages based on protected characteristic,s such as ethnic or religious grouping.
This report differs from "Data and Democracy in the Digital Age," published by the Constitution Society earlier in July, in that it doesn’t contain any recommendations, but instead sets out the possibility for nefarious activity at present and how much worse this could get in the future.
The authors, Jamie Bartlett, Josh Smith and Rose Acton, also diverged from the aforementioned report by making passing references to Facebook and no mention of Cambridge Analytica.
This publication could have been read as "the seven deadly trends" as report authors state the ways in which technology is being used in marketing, advertising and political campaigning in ways that could be perceived by some as insidious and unsettling. One gets the impression that potential voters are being viewed as potential customers by the political campaingers that are acting more like corporations.
Considerable time, effort, resource and money is being put towards capturing our attention, hearts, minds and ultimately our votes. With substantial evidence and references, Demos is demonstrating that politicians and their campaigners are crossing the "creepy line" in a bid to gain our vote.