In a post today, Facebook’s chief operating officer Sheryl Sandberg made her first public statement on a recent ProPublica investigation of ad-targeting to hate groups, calling the issue “a fail on our part.” Last week, ProPublica’s investigation found that Facebook clients could target ads using keywords like “jew hater” and “Hitler did nothing wrong.”
Sandberg claims the ad-targeting was the result of manual entries in the education and employer fields. (In simple terms, someone listed their job as “jew hater.”) That explanation contradicts the initial ProPublica article, which claimed the categories were algorithmically generated.
“We never intended or anticipated this functionality being used this way – and that is on us,” Sandberg wrote. “And we did not find it ourselves – and that is also on us.”
Sandberg laid out three changes in how the company targets ads, although each is largely an extension of existing efforts. After restricting self-reported fields for education and profession, Facebook will now restore approximately 5,000 of the most popular responses, all of which have now been reviewed to ensure they don’t violate company standards. The company will also devote more resources to ensuring that “content that goes against our community standards cannot be used to target ads,” and add more human oversight to its advertising system more broadly.
The company is also exploring new ways to encourage users to report offensive ads, although it’s unclear how those systems will improve on what is currently available. Users can already report advertising content as offensive, using the same mechanism they would for any content on Facebook. More importantly, the ProPublica investigation focused on offensive targeting rather than offensive content, and full information about how a given ad was targeted is not always immediately accessible.
In the same post, Sandberg defended the social value of ad-targeting more broadly. “A local restaurant can shoot video of their food prep with just a phone and have an ad up and running within minutes,” Sandberg wrote, “and pay only the amount needed to show it to real potential customers.”
Facebook’s ad targeting practices have come under scrutiny after news that Kremlin-linked groups spent as much as $100,000 on Facebook promotion in advance of the 2016 election. The Daily Beast has also reported that Putin-linked groups also organized 17 separate pro-Trump rallies on Facebook, a potential violation of campaign finance law. Earlier today, members of Congress sent a letter to the Federal Election Committee asking for more transparency into political ad spending on the network. “There is no reason to believe this behavior will stop in future elections,” the members wrote.
At the same time, Facebook’s ad moderation has typically been too lax to spot foreign interference at work. A Verge report this week found that the company often asks ad reviewers to examine components of ads individually, with over a thousand components to be scanned in a single day. Under those circumstances, connecting the content, targeting and buyer itself to deduce foreign interference would be almost impossible.
“I know exactly what these guys did,” one former ad-review worker said about the Russian buys. “It’s not hard to do.”