Concerns Regarding Digital Advertising Policies and Practices

The rise of the digital advertising industry has created new roles for brands, publishers, and internet platforms. It has also raised a number of concerns regarding privacy, surveillance, and fairness, accountability, and transparency around algorithmic decision-making practices.

In the early 2000s, a Minnesota father found Target ads for maternity clothing and nursery furniture addressed to his teenage daughter in the mail. According to Target statistician Andrew Pole, the company was able to use historical buying data on all of the women who had signed up for Target baby registries to identify purchasing patterns. This information was used to create an algorithm that could identify women who were likely pregnant. Target then delivered ads to women who had been identified as pregnant. For example, the company’s statisticians found that women on the baby registry were buying more quantities of unscented lotion around the beginning of their second trimester. Additionally, many pregnant women purchased supplements such as calcium, magnesium, and zinc during their first 20 weeks. By bundling pattern-based data points such as these together, Target was able to calculate and assign each shopper a “pregnancy prediction score” and estimate each pregnant woman’s due date within a narrow window.¹ In the case of the Minnesota teen, this meant that the company knew she was pregnant and acted on this knowledge before she decided to tell her own family. This was over 10 years ago.² Since then, the digital advertising ecosystem has significantly changed, and such practices have become even more refined, more pervasive, more automated, and less visible. Today, the industry relies on and monetizes user data at an unprecedented scale. In this way, data has become the lifeblood of the digital advertising industry. Simultaneously, these rampant data collection and monetization practices have raised a number of concerns.

As Harvard Business School scholar Shoshana Zuboff outlined, the digital advertising industry can be situated within the framework of “surveillance capitalism.” In her book The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power, Zuboff describes how in the twentieth century, companies such as General Motors and Ford sparked the rise of mass production and managerial capitalism. In the twenty-first century, companies such as Google and Facebook have initiated the rise of surveillance capitalism. These platforms commodify “reality” by tracking the behaviors of individuals online and offline, making predictions about how they may act in the future, and constructing mechanisms to influence these future behaviors, whether such behaviors are voting or making purchases.³

In this new digital advertising model, internet platforms might not sell access to user data. But they do sell the attention of these consumers to brands and companies that are willing to pay for it.⁴ Additionally, these platforms monetize data by using it to facilitate ad targeting. These practices are extensive and invasive. Such pervasive online and offline surveillance, and the subsequent monetization of users’ behaviors and ideas treats the “private human experience as raw material for product development and market exchange.”⁵ This model incentivizes internet platforms to collect as much data on users as possible, as this will enable them to offer precise targeting and delivery tools to advertisers.⁶ In addition, the digital advertising ecosystem incentivizes rampant data collection as these systems require vast datasets in order to operate and improve.

Despite the fact that data-driven ad targeting and delivery practices have become an integral component of the business models of many internet platforms, there is still little transparency around the mechanisms and policies that these internet platforms use to carry out these practices.⁷ Many users also lack awareness of how internet platforms target and deliver advertisements to them. Additionally, few (if any) internet platforms offer their users a comprehensive set of mechanisms with which individuals can control how they fit into the online ad ecosystem.

Nonetheless, a growing body of research has emerged documenting the concerning and often discriminatory results of algorithmic decision-making in the digital advertising ecosystem. The precise nature of modern ad targeting and delivery systems has enabled advertisers to specify which categories of users they would like to include in their ad campaigns. While these determinations might result in some users receiving ads that are relevant to them, this can also result in the discriminatory exclusion of certain categories of users. This can happen even if an advertiser sets apparently non-discriminatory parameters regarding which audiences they would like to target with an ad. This is because automated tools used to target and deliver ads at scale make inferences based on engagement metrics and data on what categories of users are more likely to engage with an ad. This can, and has been found to, reinforce certain societal biases regarding race, gender, and socioeconomic status⁸ through instances such as price discrimination in ads on online retail sites and gender and race discrimination in ads on job and employment sites.⁹ For example, an ad delivery algorithm might only deliver ads for traditionally-male dominated jobs, such as doctors or engineers, to male job-seekers—as it bases its optimization strategy on current and historical data on job occupants that reflect societal gender discrimination in these career fields. Automated tools can therefore prevent some categories of users from receiving critical and relevant information based on false inferences about who will be interested in an ad, which largely reflects societal prejudices and biases.

In addition, political advertisements on internet platforms have come under increased scrutiny. This was in part triggered by revelations that Russian operatives had used political advertising services on a number of internet platforms to influence and suppress voting in the 2016 U.S. presidential election.¹⁰ Since then, these platforms have come under increased scrutiny and pressure to provide greater transparency and accountability around their advertising operations, and to develop clearer policies and processes governing who can purchase and run political advertising campaigns.¹¹ Some platforms, such as Pinterest and most recently Twitter, have opted to ban political advertising on their platforms altogether.¹² However, because political advertisements cannot be easily categorized or defined, and given that some of these decisions were made fairly recently, the impact of these new bans is yet to be seen.

Policymakers are also considering what approaches they can take to prevent election interference in the future, and to promote greater transparency around political advertising practices online. One example of such an effort is the introduction of the Honest Ads Act, which aims to regulate political advertisements online. However, some commenters have questioned whether such legislative approaches are consistent with the First Amendment, and whether this is a space in which the government can legally take action.¹³