Society problems

Facebook Ad Algorithms Can Harm Democratic Society

Piotr Sapiezynski, a researcher at Northeastern University, recently told the European Parliament that Facebook’s ad serving algorithms could be detrimental to both political campaigns and society as a whole.

Sapiezynski testified at a hearing on a bill regarding the transparency and targeting of political advertising in Brussels.

“Facebook’s algorithms make it harder and more expensive for political parties to reach potential voters who Facebook believes don’t already agree with the advertised message,” said Sapiezynski, associate research fellow at the Cybersecurity and Privacy Institute. du Khoury. College of Computer Sciences, during a public hearing of the European Parliament’s Internal Market and Consumer Protection Committee on July 11.

The proposed regulation on the transparency and targeting of political advertising would be the first to govern political advertising in the European Union online, as current legislation only governs traditional media such as television and newspapers.

Piotr Sapiezynski, a research associate at the Khoury College of Computer Sciences, says tech companies should be required to give users, journalists and researchers access to the information needed to hold both advertisers and platforms accountable. Photo by Matthew Modoono/Northeastern University

The draft law provides general transparency obligations for all actors involved in the financing, preparation, placement and dissemination of political advertising, offline and online, and aims to protect the personal data of individuals by establishing rules on the use of technical ad targeting and amplification. The proposal aims to increase the transparency of political advertisements across the European Union ahead of the next European Parliament elections in 2024.

Sapiezynski and his collaborators have been studying Facebook’s advertising platform for a few years, he said, finding that the platform’s advertising algorithms demonstrate gender, racial and other biases. Their findings about Facebook’s practices for collecting and using personal data without consent were part of a $5 billion settlement with the US Federal Trade Commission.

They also identified that Facebook’s algorithms discriminate on the basis of gender and race in serving job and housing ads, for which the US Department of Justice recently charged Facebook’s parent company Meta. Platforms Inc. in June in the first case challenging algorithmic discrimination under the Fair Housing Act.

“We assume it’s a for-profit business, so they do what makes the most money,” Sapiezynski said. “It turns out that maximizing profit does not align with the goals of a healthy society.”

In their studies, Sapiezynski et al. took a closer look at ad serving versus ad targeting. Facebook gives advertisers many tools and options to analyze and select target audiences who will see the ad based on demographics, location, interests, and more, Sapiezynski said. Advertisers may display micro-targeting or discriminatory behavior during this phase; However, the researchers are most concerned about the harms that arise when Facebook makes certain algorithmic decisions in the second phase, ad serving, even when the advertiser is well-intentioned or uses broad targeting criteria.

“The budget the advertiser sets for a particular ad is usually not enough to reach every person in the target audience,” Sapiezynski said. “The platform decides who among the targeted audience will actually see the ad.”

The researchers found that the delivery decision is not made at random. It’s made by a profit-optimized algorithm based on all the user data Facebook has, Sapiezynski said.

The more time a user spends on Facebook, the more ads they see and the more money Meta earns. Ad serving algorithms are designed to maximize the relevance of ads to users and avoid showing them disturbing or uninteresting ads in the hopes that users will continue to browse and view additional ads.

“They avoid showing uninteresting ads, or they charge more to show them to cover potentially lost revenue,” Sapiezynski said.

In one of the studies Sapierzynski and his colleagues conducted in the United States in 2019, they ran ads for then-presidential candidates Bernie Sanders and Donald Trump, targeting Democratic and Republican audiences. Although in the first experiment the ads targeted the same general audience, ads for Bernie Sanders were shown primarily to Democrats, while ads for Donald Trump were shown primarily to Republicans, Sapiezynski said.

In other experiments, researchers found that it was up to four times more expensive to serve ads to the targeted audience that already disagreed with the message and did not support the featured candidate.

“In this way, Facebook limits the ability of political advertisers to reach audiences that Facebook believes do not share those advertisers’ political views,” Sapiezynski said. “This is in stark contrast to advertising in traditional media, where reaching a voter costs the same, regardless of the identity of the political advertiser.”

People are more likely to see ads that reinforce their worldview and less likely to be exposed to messages from other parties, even if those parties are actively trying to reach various Facebook users. Users who are algorithmically blocked from seeing different advertisements and political viewpoints will have limited sources of information when making their choices in the voting booth, Sapiezynski said.

“This optimization plays a significant role in defining which voters see which political messages, and it is inconsistent with the goals of a well-informed democratic society. If the legislation focuses solely on limiting advertising targeting options, it will give advertising platforms even more power to make non-transparent serving decisions, and it will not ensure meaningful improvements in the experience of real users” , Sapiezynski said.

He called for legislation that explicitly recognizes the distinct roles that advertisers and platforms play and that will require tech companies to give users, journalists and researchers the information needed to hold advertisers and platforms accountable.

Facebook has already been forced to post all political ads to the Facebook Ads Library, Sapiezynski said; however, it does not contain enough information to understand what is really going on.

“All we can see at the moment is what kind of posts politicians are advertising, but we don’t fully understand who they are targeting with these posts, nor do we understand what Facebook is doing with them. those targeting options, how Facebook optimizes delivery,” he said. “We only get a rough estimate of the breakdown of people who saw the ad by gender, age, and state.”

At the heart of it all is the handling of personal data, Sapiezynski said, because Facebook’s algorithms work with “millions” of sensitive characteristics that define or describe a user. Some of the supporters of the new European transparency regulation are pushing to drastically limit the ability of online platforms to process personal data. They propose allowing platforms and advertisers to use a short list of features and prohibiting them from targeting users with other features.

“I am ready to believe that he [Facebook advertising algorithm] works well for products. The problem really arises when you think about higher stakes, advertising life opportunities, etc. Sapiezynski said. “And what we’re seeing is that Facebook is essentially replicating the biases that we already know exist.”

For media inquiriesplease contact media@northeastern.edu.