Why we're concerned about profiling and micro-targeting in elections.

Political campaigns around the world have turned into sophisticated data operations. They rely on data- your data- to facilitate a number of decisions: where to hold rallies, which States or constituencies to focus resources on, which campaign messages to focus on in which area, and how to target supporters, undecided voters, and non-supporters.

While data driven political campaigns are not new, the granularity of data available and the potential power to sway or suppress voters through that data is. The way in which data is used in modern political campaigning is highly privacy invasive, raises important security questions, and has the potential to undermine faith in the democratic process. The increased reliance by major political parties to profile and micro-target people online for political messaging is of great concern to Privacy International.

For example, it is well known that Facebook’s business model is so lucrative because of the ability to offer targeted advertising, based on user information like age, location and interests. This has proved so appealing that political parties want in. In the same way that online advertising targets people based on interests, personality and mood to ultimately sell products, political parties try to persuade you to buy what they are selling come election time.

There is a complex and opaque corporate ecosystem behind targeted online political advertising.  This isn't just the Facebooks and Twitters of the world - data brokers and data analytics companies should all be part of this conversation.  Data analytics and digital media firms are employed directly by political parties contesting elections to run online campaigns. The details are often unclear- exactly who these companies work for, what they do and how they do it is often a guarded secret. What is clear is that there are thousands of companies whose business model it is to exploit the data people share online in such a way that intimate personal details about a person’s beliefs, habits, and behavior can be better understood and used for the purpose of allowing political parties to target these individuals with political messages.

Particularly in countries where there is history of political violence, campaigning based on intensive collection of personal data is untested ground fraught with great risk. Collecting data on ethnicity or political affiliation with no limitations or safeguards is open to abuse. Plus, many countries still lack sufficient laws and regulatory mechanisms to safeguard data protection and privacy affected by this level of data generation and processing, particularly sensitive personal data such as political views or ethnicity.

What are targeted messages?

The existing data and advertising ecosystem targets content and adverts to specific audiences. This is commonplace, and allows companies to not just reach very specific people, but also at specific times and places.

For political campaigns, this has several unique advantages: Unlike posters or broadcast political ads, which are public by default, campaigns can show different ads, and different content to different people. This means that they can make different promises or even contradictory claims with little oversight or accountability. This has taken political advertising out of the public sphere- we can all see posters/billboards on the street or adverts on the television and to some extent hold political parties accountable for their messaging and how much they spend. When only some people see a message, or everyone sees a different one, this makes the process less accountable.

Another unique advantage for political campaigns is the ability to reach very specific groups of people, such as undecided voters, or voters in marginal constituencies. This means that campaigns can strategically discourage some people from voting at all, be confident which promises will be most persuasive for others, and spread misinformation or incite fear in those that are most vulnerable.

What is Profiling?

Profiling is a term outlined in the European General Data Protection Regulation (GDPR). It is defined as “any form of automated processing of personal data consisting of the use of personal data to evaluate certain personal aspects relating to a natural person, in particular to analyse or predict aspects concerning that natural person’s performance at work, economic situation, health, personal preferences, interests, reliability, behaviour, location or movements”.

In short, this means that organisations, many you've never heard of, are able to learn about your habits, personality, sexual interests, political beliefs, and more to make predictions about your personality and behavior. This is true even if you have not shared this information with them. This is especially concerning when sensitive information, such as political beliefs or personality traits are inferred from completely unrelated data using profiling.

Cambridge Analytica is just one example of a company that obtained data through harvesting and profiling, and used these very detailed profiles to target voters. Other campaigns rely exclusively on the tools that platforms make available. For instance, Facebook offers a tool called LookalikeAudiences, that automatically finds and refines audiences that are similar to those who visited a pages’ website, or liked that page. This is reportedly how the AfD in Germany  - a relatively new political party - was able to strategically reach out to a diverse set of people that are similar to their existing voter base in one way or another.

Voter profiling and targeting doesn’t end online - campaigns in the US have used postcode level data to hone their messaging to voters on their doorstep.

What next?

There needs to be much more transparency about the way our data is collected and used by political parties and the companies they contract. Political parties should be transparent about the campaigns they have funded, how they have developed targeted messages, or which companies they have worked with.

Where data is generated, individuals should be able to find out which companies hold what kinds of data about them. Profiling generates highly sensitive inferences and predictions about people’s personality, behaviour or beliefs. Individuals should be able to access these inferences and predictions about them, in order to effectively challenge them, or to ask for profiles to be deleted.

For more information on how PI is tackling this issue, see our Learning page on Data and Elections.