Data as Political Intelligence

The digital campaign led by Brad Parscale for the 2016 Trump presidency reportedly tested 40,000 to 50,000 versions of messages per day. This level of testing not only led to messages that demonstrably moved people to action, like donations, but also led to insights into what voters were motivated by and what they wanted to hear, which the campaign adapted per individual or per group.

Third-party trackers embedded in such messages or on websites stay with a user long after the original site is closed, leading to much further data collection than a user might imagine. Campaign apps at times went one step further by not only collecting data and testing in real time, but also by motivating users to provide more information and even reach out to their own contacts.

Similarly, the use of methods such as digital listening—collecting openly available information on political discussions—leads to the accumulation of insights into voter opinions, which can then be used to form positions, decide which areas to campaign in, or how to pitch a speech to a certain community.

Gathering information about voters and fine-tuning a message for a certain audience are not new phenomena. However, the ways in which information is being collected on unsuspecting participants are specific to the Influence Industry and are not being open discussed and debated. Likewise, the messages being tested and subtly altered to appeal to specific voters are occurring at an unimaginable scale. The social and political impact of ever-increasing personalisation in digital advertising and communication, which can lead to skewing voters’ understanding candidates priorities and agenda, needs to be considered.

A/B testing

Since it was first used in the 1920's, A/B testing has been integrated into campaigning and has become part of standard campaign practice for websites, emails (subject lines, bodies), design elements (images, background buttons), headlines, direct mail, TV, radio, phone calls and even texting to "find the right messaging." A/B testing is now standard practice among virtually any entity with an online presence.

As A/B testing services become more automated, algorithms can create far more variants and combinations of text, media, buttons, etc. based on campaign inputs. This ostensibly means that machines - instead of people - would decide what a potential voter reads and sees, which could set a precedent of creating personalised political content free of human oversight.

In practice

The pro-Brexit Vote Leave campaign split voters into three groups: those firmly voting to remain, those voting to leave, and those on the fence. Vote Leave invested 98% of its marketing budget in digital efforts focused on this third group and tested five narratives on them. The winning message was "take back control".

A screenshot from the RNC Testing Booklet shows how Donald Trump’s campaign tested these two background images against each other on its donation page. The image of Trump performed about 80% better than the image of Clinton.

Two images tested against each other on Donald Trump's donation page, Source: 2016 RNC Testing Booklet

NGP VAN, 'MiniVAN & MiniVAN Manager', YouTube, 2016, Source: https://www.youtube.com/watch?v=0qvF3C-iqCA

Psychometric profiling

The simplest option in conducting this type of profiling is to conduct a survey in which individuals answer questions that reveal aspects of their psychological composition.

Today, direct user input is not necessary anymore, as researchers have claimed that personality traits can be predicted from a user's "Like" activity on Facebook or other public social media interactions and data traces.

In practice

Famously, Cambridge Analytica identified and targeted voters in the run-up to the 2016 presidential election for candidate Ted Cruz, and later for Donald Trump. Former CEO Alexander Nix explicitly linked the company's targeting of personality traits with influencing voting behaviour:

See for yourself

A Drop in the Ocean, a tool that enables visitors to take their own personality test, which then feeds them political ads.

Check out the app (link opens in a new tab): A Drop in the OCEAN

Third-party tracking

In the run-up to an election, a voter may want to research candidates by visiting their affiliated websites and social media presences. Many political campaigns specifically promote a ‘political cookie’, which is a piece of data that can be used to match a person’s online identity with their offline details.

For example, during Colombia's 2018 national election, an analysis of websites belonging to leading candidates revealed extensive use of third-party tracking tools. Of the leading 21 candidates' websites, eight had third-party Facebook trackers, 12 had Twitter trackers and 11 had some form of tracking on the donation page. Among 10 political party websites, five had Facebook trackers, seven were from Twitter, and five had other trackers on their donation page.

Even when candidates and parties are different, these sites could be using the same tracking companies to monitor activities like donations, newsletter signups and clicks. The data can be collected using a variety of tools that are often implemented without the users' knowledge.

Importantly, a tracking company could then cross-reference this browsing activity and combine it with external data sources to profile the user.

See for yourself

To see these tools in action, try the Electronic Frontier Foundation's tool Cover Your Tracks to discover how trackers are used on your browser.

Digital listening

Data is gathered through scraping or trawling software from public posts, such as tweets or Facebook posts connected to hashtags or keywords.

This data is analysed to infer different pieces of information, such as whether a tweet demonstrates a positive or negative sentiment by noticing the words and contexts in which they appear.

In 2019, our team spoke with industry professionals about their work, one of whom, speaking under anonymity, stated: We don’t do questionnaires. We basically track data, everything people posting in a public forum and we arrange those simply around a specific keyword or a topic. Because we’re not asking people, introduce your attitudinal bias. So we just passively observe what people say naturally. I know how you can manipulate a poll question to give whatever answer you want. So we simply just observe. We don’t even bother asking questions, and that’s where it gets really scary because there was a service where you can track somebody’s like, 70 of their previous likes. You can deduce their attitude about social issues.

In practice

In Taiwan, a first-time candidate running for Mayor of Taipei, Ko Wen-je worked with the firm AutoPolitic.

For Dr. Ko's campaign, the company measured public sentiment to understand 'what topics the public cared about (and why), who the influencers are (so they can engage them) and what topics the influencers are most interested in'.

They concluded that Dr. Ko should engage with young people through activities including tattoos, street dancing, basketball and riding bicycles. Dr. Ko followed this analysis and his visit to a tattoo parlour was considered successful, as it was shared widely on social media platforms.