ABSTRACT
Recent attention paid to 'dark patterns' in online marketing has yet to engage with the most prevalent forms of ad targeting online. The dominant model we have chosen for supporting the online information environment is based on data-driven targeted advertising. These targeted ads are 'dark' in the sense that they are visible only to those to whom they are directed, which poses challenges for public accountability. Their sheer quantity is matched by their ephemerality. They are also 'dark' in the sense that the overall pattern of their delivery is invisible. Not only can we not see the ads directed to others, but we cannot see who else is being targeted the way we are. These 'dark patterns' raise three sets of social concerns about discrimination and socially detrimental forms of messaging. First, dark patterns of advertising can reproduce historical forms of discrimination against protected groups. Second, they may enable the circulation of forms of stereotyping that fly under the radar of public accountability. Third, they enable new forms of granular discrimination that, even if legal, exacerbate social division and injustice. For example, in the United States, dark ads have been used to target and suppress the voter turnout in African-American communities in swing states. Dark ad patterns have also been used to target job ads to particular age groups in violation of US anti-discrimination law. In this case, the individual ads are not the problem – rather the detrimental consequences are a result of their overall pattern of their distribution. With these social concerns in mind, this article reports on experiments designed to submit targeted advertising to public scrutiny.
