Facebook advertisements have enabled discrimination based on gender, race and age.

Melbourne

Social media platforms are transforming how online advertising worksand, in turn, raising concerns about new forms of discrimination and predatorymarketing. Today the ARC Centre of Excellence for Automated Decision Making andSociety (ADM+S) — a multi-university entity led by RMIT — launchedthe Australian Ad Observatory. This research project will explore howplatforms target Australian users with ads. The goal is to foster aconversation about the need for public transparency in online advertising.

The rise of 'dark ads' In the mass media era, advertising was (for themost part) public. This meant it was open to scrutiny. When advertisers behavedillegally or irresponsibly, the results were there for many to see.

And the history of advertising is riddled with irresponsible behavior.We've witnessed tobacco and alcohol companies engage in the predatory targetingof women, underage people, and socially disadvantaged communities. We've seenthe use of sexist and racist stereotypes. More recently, the circulation ofmisinformation has become a major concern.

When such practices take place in the open, they can be responded to bymedia watchdogs, citizens, and regulators. On the other hand, the rise ofonline advertising — which is tailored to individuals and delivered on personaldevices — reduces public accountability. These so-called "dark ads"are visible only to the targeted user. They are hard to track, since an ad mayonly appear a few times before disappearing. Also, the user doesn't knowwhether the ads they see are being shown to others, or whether they are beingsingled out based on their identity data.

Severe consequences There's a lack of transparency surrounding theautomated systems Facebook employs to target users with ads, as wellas recommendations it provides to advertisers. In 2017 investigativejournalists at ProPublica were able to purchase a test adon Facebook targeting users associated with the term "Jewhater". In response to the attempted ad purchase, Facebook's automated systemsuggested additional targeting categories including "how to burnJews". Facebook removed the categories after being confrontedwith the findings. Without the scrutiny of the investigators, might they haveendured indefinitely? Researchers' concern about dark ads continues to grow. Inthe past, Facebook has made it possible to advertise housing, credit,and employment based on race, gender, and age.

This year it was found delivering targeted ads for military gearalongside posts about the attack on the US Capitol. It also enabled ads targeting AfricanAmericans during the 2016 US presidential campaign to suppress voterturnout. Public support for transparency It's not always clear whether suchoffenses are deliberate or not. Nevertheless, they've become a feature of theextensive automated ad-targeting systems used by commercial digital platforms,and the opportunity for harm is ever-present — deliberate or otherwise.

Most examples of problematic Facebook advertising come fromthe United States, as this is where the bulk of research on this issue isconducted. But it's equally important to scrutinize the issue in othercountries, including in Australia. And Australians agree. Researchpublished on Tuesday and conducted by Essential Media (on behalf ofthe ADM+S Centre) has revealed strong support for transparency in advertising.More than three-quarters of Australian Facebook usersresponded Facebook "should be more transparent about how itdistributes advertising on its news feed".

With this goal in mind, the Australian Ad Observatory developeda version of an online tool created by ProPublica to let members ofthe public anonymously share the ads they receive on Facebook withreporters and researchers. The tool will allow us to see how ads are beingtargeted to Australians based on demographic characteristics such asage, ethnicity, and income. It is available as a free plugin for anyone toinstall on their web browser (and can be removed or disabled at any time).Importantly, the plug-in does not collect any personally identifying information.Participants are invited to provide some basic, non-identifying, demographicinformation when they install it, but this is voluntary. The plug-in onlycaptures the text and images in ads labeled as "sponsored content"which appear in users' news feeds.

Facebook's online ad library does provide some level of visibility intoits targeted ad practices — but this isn't comprehensive. The ad library onlyprovides limited information about how ads are targeted and exclude some adsbased on the number of people reached. It's also not reliable as an archive,since the ads disappear when no longer in use. The need for public interestresearch Despite its past failings, Facebook has been hostile towardsoutsider attempts to ensure accountability. For example, it recently demandedresearchers at New York University discontinue their research intohow political ads are targeted on Facebook. When theyrefused, Facebook cut off their access to its platform. The techcompany claimed it had to ban the research because it was bound by a settlementwith the United States' Federal Trade Commission over past privacyviolations. However, the Federal Trade Commission publicly rejectedthis claim and emphasized its support for public interest research intended"to shed light on opaque business practices, especially aroundsurveillance-based advertising". Platforms should be required to provideuniversal transparency for how they advertise. Until this happens, projectslike the Australian Ad Observatory plugin can help provide someaccountability. To participate, or for more information, visit the website.