![Here's how social algorithms influence elections A study co-signed by the Polytechnic University of Milan explains](https://data2.nssmag.com/images/galleries/39591/thumbs/cover-elections-underwood.jpg)
Here's how social algorithms influence elections
A study co-signed by the Polytechnic University of Milan explains
July 20th, 2024
A new study published in the journal PNAS Nexus reveals how social media algorithms favor political content sponsored by certain parties, given the same investment budget. This research, born from a collaboration between the Politecnico di Milano, the Ludwig Maximilians Universität of Munich, and the CENTAI institute of Turin, analyzed over 80,000 political ads on Facebook and Instagram before the 2021 German federal elections. The political ads in question were published by parties across the political spectrum and generated over 1.1 billion impressions from more than 60 million eligible voters. The study examined the effectiveness and reach of these ads, highlighting substantial inequalities in online campaign results. Most notably, significant discrepancies in the effectiveness of the ads and the intensity with which they reached their targets were observed, favoring extremist groups.
@theicedcoffeehour Facebook Stole All Your Data #cambridgeanalytica #markzuckerberg #metaverse original sound - The Iced Coffee Hour
One of the fundamental discoveries was that over 70% of political parties used user profiling in their ads, targeting specific demographic groups. Essentially, this practice of microtargeting, which became famous during the Cambridge Analytica scandal where Facebook data was used to create tailored political ads, significantly influenced the efficiency of political campaigns, now turned into battles fought across the web. Calculating the variations in advertising costs, measured in impressions per euro spent, showed that not all parties achieved equal results with the same budget. The far-right Alternative für Deutschland (AfD) proved to be the most efficient, with their ads being almost six times more effective than those of other parties with the same budget. In contrast, the Green Party was the least cost-effective. Francesco Pierri, a researcher from the Data Science research group of the Department of Electronics, Information, and Bioengineering at the Politecnico di Milano, who co-led the study, suggests that the success of ads from populist parties like the AfD can be attributed to the polarizing, often sensationalistic nature of their incendiary political content, which algorithms tend to favor. This algorithmic preference means that certain parties' ads receive disproportionate reach and engagement, thus reinforcing their political messages more effectively than others.
The study also discovered significant discrepancies between the expected demographic targets and the actual audiences reached. Most parties tended to reach a younger demographic than expected, while far-right parties reached an older audience than expected. Pierri and his colleagues hypothesize that algorithmic bias in ad distribution is based on known voter behaviors and their interactions with social media content. The systematic bias in ad distribution not only distorts the political landscape but also limits the political participation of disadvantaged groups. Higher costs for similar ads imposed on certain parties can harm fair political competition. Pierri emphasizes the need for greater transparency from social media platforms regarding their political advertising practices to ensure fair elections.
The implications of this study are far-reaching, as they not only raise concerns about political promotion practices but also highlight the lack of effective monitoring and regulation of political advertising on social media to safeguard the integrity of democratic processes. The EU's Digital Services Act, for example, aims to improve transparency and accountability in digital advertising, and public pressure has also led social media platforms to provide greater access to political and social ads, enabling large-scale studies like this one. These efforts are crucial for understanding and mitigating biases and beliefs disseminated worldwide through algorithmic ad distribution. The study underscores the urgent need for transparency and fairness in political advertising on social media platforms – suggesting (and this is our addition) that currently, this transparency and fairness are lacking. The final message of the study, in short, is that marketing should be recognized as marketing and not allowed to unduly influence, and by learning to recognize and address its effects we can begin to work towards more equitable and democratic electoral processes. The collaboration between leading research institutions highlights the importance of interdisciplinary efforts in tackling these complex challenges and advancing our understanding of digital influence on politics.