Mystery around the methods online platforms use to target users

Published: 18 February 2020

People need the power to decide on how they want to protect themselves” – Katarzyna Szymielewicz, co-founder and president of Panoptykon Foundation.

Political campaigns increasingly target voters online with messages appealing to personal characteristics, a strategy known as political microtargeting (PMT). A Civitates-funded project by watchdog NGO Panoptykon Foundation sought clarity on the scale and impact of PMT in Poland, gathering evidence to inform European policy makers. Panoptykon Foundation is a non-profit that protects fundamental rights and freedoms in the context of fast-changing technologies and growing surveillance. The organisation engages in research, awareness-raising, civic engagement and advocacy towards policy makers.

Can you describe to us what the project is about?
The role of political microtargeting (PMT) in the 2016 US elections and the Brexit vote has raised public alarm about the use of personal data and the repercussions for democracy. Despite this concern, a lot of mystery remains around the methods online platforms use to target users. We want to give people the power to decide on how they want to protect themselves. For that we need to explore what is going on, so we used the Polish parliamentary elections in autumn 2019 as a main testing ground to understand the risks it represents for public discourse in Europe. To this end we joined forces with Who Targets Me, which developed a browser plug-in used to gather data on who is targeting users.

To what extent is transparency available?
The global platform advertising market is mainly divided between two giants: Facebook and Google. Both platforms have made efforts to meet demands for transparency, launching political advertisement databases showing targeting criteria including language, demographics, basic interests and location. Panoptykon Foundation found that these archives capture the scope of political advertising quite well. However, deeper insight into how custom audiences are selected is still missing. The user gets to see the tip of the iceberg when it comes to how they are targeted.What remains out of view are the big data analytics performed by online platforms: exactly what user data is analysed and how.

What are your conclusions so far?
Most analysis of PMT until now has been rooted in the US context. However, there are key differences between the US and European markets influencing advertising behaviour. While in the United States there is a large market for paid political advertisements, in Europe we found out that the big thing is organic content, which can also be targeted, especially by excluding certain audiences. Political ads in Poland did not have a strong tendency towards polarising messages or attack campaigns. Political messaging in organic (unpaid, user-generated) content, on the other hand, tended to be more polarising and “dirty.” Creating and spreading this content isn’t a simple case of free speech, as it isn’t known how platforms select which content gets a boost and to whom it becomes most visible. This is where the European and American contexts converge: both lack insight into how platforms are shaping organic engagement.

What could according to you be a solution?
We envision a clever mix of regulation and self- or co-regulation of platforms to reign in voter manipulation through PMT. We are pushing for the same transparency on targeting criteria to be available to both paying political advertising clients and non-paying users. To get the full picture of how PMT is influencing the political landscape, it is crucial that transparency measures expose the way platforms analyse user data and select custom audiences. Insights provided by plug-ins like the one developed by Who Targets Me are in the end dependent on the accessibility of platform data.

Is there anything within this project that makes you feel really proud?
I am very proud of our small team at Panoptykon as we handled a very complex project, involving a lot of coordination and new types of activities, especially data analysis. We have learnt a lot in the process of collecting and analysing data. I think it takes a lot of courage for lawyers to take such a steep learning curve.