Targeted Ads

This report summarizes the findings of a one-year study of online targeted advertisements. It highlights important questions and considerations regarding how online targeted ads uphold, produce, and recreate racially discriminatory infrastructures within everyday life. First, we propose a novel framework–algorithmic discrimination–which purports targeted ads to be discriminatory infrastructures by design: namely, this conceptual and analytic tool situates the potential harms and risks of targeted ads in relation to a longer history of predatory processes, tactics, and classification schemas, especially within and against marginalized communities of color. Next, we discuss how this framework relates to our novel methodology for algorithmic discrimination audits, in light of ongoing discussions of algorithmic accountability and corporations’ seeming attempts to forestall such efforts. Focusing on third-party search data pertaining to queries for educational opportunities, employment, and housing, we use zip codes as a proxy for racial and sociodemographic data, to audit and assess trends in online ad targeting. We compare differences across and within neighborhoods in online targeting patterns; we also compare individual ad messaging content. In contrast, we argue that a sociohistorical and infrastructural approach to algorithmic audits elucidates the community-based harms and risks of targeted ad systems as well as the digital infrastructures targeted ads undergird and fuel. As such, this approach more aptly shows the longer-term impacts of targeted ads and how they re-instantiate–and amplify–legacies of racial inequality. We close with key questions and future directions for this exploratory framework and methodology, particularly considering ongoing concerns about tech regulation and policy, and the protection of vulnerable communities from further tech-driven exploitation and extraction.