Can a smart #AI strategy convince EU that #Facebook content moderation is up to snuff?

On 17 February, while meeting with European policymakers in Brussels, Facebook CEO Mark Zuckerberg (pictured) released a white paper outlining his strategy for dealing with harmful content on social media platforms. In response to the European Union’s desire to ratchet up oversight of the tech sector, the billionaire made a case that the industry should be regulated at a global, rather than a local, level.

He also argued that tech companies’ accountability for harmful content on their platforms should be limited—as long as they have effective content moderation systems in place. Zuckerberg’s suggestions, however, received a decidedly frosty reception in Brussels. Even after meeting Zuckerberg, the EU maintained that Facebook’s new content moderation proposal fell short on many fronts, particularly failing to delineate the concrete steps the platform would take to combat illegal content online. “It’s not enough. It’s too slow, it’s too low in terms of responsibility and regulation,” opined data strategy commissioner Thierry Breton after his sit-down with Facebook’s top executive.

Where is Facebook’s current system falling short?

Tech giants have a fairly narrow window in which to convince European policymakers that their own efforts to curb problematic content are sufficient. The Commission’s Digital Services Act, the long-awaited revision of the EU’s 20-year-old e-commerce directive, is slated to be rolled out later this year. As Breton noted, “if we see that [Facebook’s own process for moderating content] is not what we need regarding our own standards, we will have to regulate and put this in our [Digital Services Act]”.

Facebook is determined that it is able to solve the problem of content moderation itself—by using the advancements in Artificial Intelligence (AI) and machine learning to boost its current system. The thorny task of identifying and removing problematic material is currently outsourced to third-party professionals who use their discretion to determine why an image or video is objectionable.

Reliance on such human oversight is problematic given the gruesome nature of some content moderators are forced to sift through—the fact that employees moderating content for Facebook in Lisbon, Dublin and Warsaw have attested that their work has given them PTSD is sure to raise red flags in the European institutions. In addition, the efficacity of human moderation has been called into question after Facebook has repeatedly failed to quickly remove harmful content such as the 2019 live streaming of the Christchurch mosque attack.

AI can plug some of Facebook’s blind spots

Artificial intelligence has the potential to significantly decrease such dependence on human moderation. Algorithms trained by machine learning can support moderators by not only reducing their exposure to harmful online content but also by augmenting their productivity and accuracy. Facebook claims to have seen promising results in its experiments with AI so far—according to the platform, during the second half of 2019 the platform managed to flag 99.8 % of graphic content and 99.5% of nudity and child exploitation even before users signalled them out. Of course, the technology still needs finetuning—content moderation algorithms based on AI still particularly struggle with recognising the nuances of hate speech, for example— an issue the European Union is especially concerned by.

What’s more, the European Union is wary of allowing Facebook a free reign in exploiting such a sensitive technology as AI. These concerns primarily stem from Facebook’s poor track record in meeting Europe’s transparency standards—raising the spectre that the company might develop ‘black-box algorithms’ preventing third-party or government audits on the platform.

At the heart of the digital policy— released last week by the EU’s digital czar and competition chief, Margrethe Vestager—is regulating how companies use artificial intelligence. While the EU wants governments to monitor the riskier aspects of it, such as the use of facial recognition technology, many apps in Europe’s own backyard are demonstrating how it can be used both safely and effectively. Can Facebook take a cue from them?

Next gen social media apps make their mark by using AI innovatively

One European app in particular has creatively harnessed artificial intelligence to provide its young users with a safer and more pleasant experience. Yubo, a live-streaming app for ages 13-25, relies on a balance between complex algorithms and educating young people about online harms to keep its 20 million users safe on the platform. A carefully trained algorithm detects problematic content such as nudity or someone only in their underwear in a live stream—which is when Yubo’s real-time intervention system kicks in, sending a warning message to users explaining why such behaviour is objectionable. If the user doesn’t correct their behaviour, the stream can be shut down remotely.

In order to make sure that the app is reserved to teenagers and young adults, meanwhile, Yubo has partnered with the popular UK-based digital identity provider, Yoti, which matches uploaded identification documents to the user’s selfies using facial recognition software. Yoti’s technology—which is also being deployed in supermarkets to make sure that under-18s aren’t buying restricted goods like cigarettes or alcohol— can be employed by other social media apps to “age-gate” their community or to make sure users are who they say they are.

Yubo’s attempts to nip problematic online content in the bud—by combining computer-assisted moderation with sensitizing young users about the do’s and don’ts of good online protocol—have made a splash in the tech sector. They should also provide an encouraging example to both tech giants like Facebook and regulators such as the European Union that social media platforms can strike a healthy balance between being overly reliant on either AI or human oversight.

Artificial intelligence isn’t a cure-all for the challenges which social media platforms face. As Facebook seeks to convince the EU that it’s able to moderate its own content, however, it would be well-served by taking a cue from the new generation apps which have managed to develop a holistic strategy, incorporating AI, for keeping their users safe.

Share this article: