Three Ways to Clean Up the Toxic Minefields of Social Media

Social media is not just a communication space, but a new kind of battle space. And although the forces of good are beginning to pay more attention to this problem, the hard reality is that they are still losing. 

A toxic witch’s brew of disinformation, conspiracy theory, and hate continues to plague the platforms through which much of our modern information networks run, hitting literally every important issue. Public health professionals say our fight against the coronavirus pandemic has been rendered far more difficult by the accompanying “infodemic.” A recent Oxford University study of disinformation concerning the disease and government responses documented 225 distinct conspiracy theory campaigns, 88 percent of which used social media as their hub. Meanwhile, protests over systemic racism and injustice continue to be warped by online efforts to push false stories and conspiracy theories. And that doesn’t even begin to talk about the historic levels of disinformation surrounding the upcoming U.S. election. 

Recent actions taken by platforms like Facebook and Twitter to ban political ads, demonetize hateful content, and put disclaimers on certain types of content that is broad reaching and verifiably false, show that social media companies can, in fits and starts, make changes for the better. 

Yet their approach is woefully late and lacking. It is often characterized by an ex-post response, lack of sustained action, and insufficient sense of urgency. Misinformation about the western wildfires, for example, had already gone viral by the time the companies began to try to rein it in. And Facebook’s much-ballyhooed oversight board has announced that it would likely not be operational in time for the 2020 election. Civil rights organizations remain rightly appalled at the situation, while the firms’ own employees lament in leaked internal documents that “We are failing. And what's worse, we have enshrined that failure in our policies.”

Professionals who work on digital threats worry that the problem will only grow worse in the coming months. The Russian efforts from 2016 never really ended, and many others have since joined in. Perhaps most significantly, as the pandemic took hold, the Chinese government drastically accelerated its information warfare activities this year. There are already efforts by both foreign governments and domestic extremists to use social-media platforms not just to muddy the upcoming election, but to spur violence during its fraught aftermath. And when or if we do get a COVID vaccine, its rollout will be targeted in much the same way. As the chairs of the bipartisan U.S. Cyberspace Solarium Commission concluded, the battle against disinformation is not just about protecting our democracy. It is now of "truly life or death importance."

The authors have collectively wrestled with the problem of disinformation campaigns and violent hate speech from experiences of researching and working on the issue with the tech industry, the military and government, and the human rights community. We know that there is no easy remedy to this problem — for three primary reasons.

One is the sheer scale of the challenge. For instance, based on research enabled by Twitter’s streaming API, we’ve found that on any given day, between 0.5 percent to 1.5 percent of all tweets qualify as violent or hate speech, which equates to hundreds of millions of posts a year that are encountered by hundreds of millions of users. There are similar problems on the other platforms. On Facebook, websites promoting coronavirus conspiracy theories have more than ten times the engagement as public health organizations. 

Second is the very design of the platforms themselves, which tie their profit to our psychology. In a system that monetizes clicks, hate speech fueled by disinformation can quickly gain attention and engagement for the very reason that it confirms pre-existing biases and plays out in a structure that rewards trends. 

Finally, there are multiple competing priorities. One is how private profit and the public good simply aren’t always aligned. As a Facebook whistleblower described their attempts to do the right thing while policing dangerous content: “I consider myself to have been put in an impossible spot – caught between my loyalties to the company and my loyalties to the world as a whole.” 

Yet the users also send competing signals. They simultaneously want to see less offensive and false content, but don’t like platforms making decisions on censorship. They don’t want companies deciding the truth for them, but are even more opposed to the idea of the government making such determinations.

 With paralyzed domestic politics, and a dose of ignorance on tech issues, don’t expect Congress to solve these issues anytime soon. Still, our collective experience hints at a better way to mitigate at least part of this mess. Here are three principles that could help social media firms to reduce the amount of false and hateful information encountered and consumed, while still preserving what is loved, and profitable, about the online world.

“If you can’t measure it, you can’t manage it.” As long as there are people and the internet, there will be attempts to misuse it. Yet, as business guru Peter Drucker would advise, we’ll never be able to manage this problem unless we can solve the ironic problem of lacking data about this data-fueled space. Most social-media platforms now finally do some self-reporting on false and abusive account and content “take-downs,” but these reports are often intermittent and usually inconsistent across different platforms, including even ones owned by the same company. Users, advertisers, and regulators have no real way of knowing whether the problems are getting better or worse, or whether the platforms are devoting significant resources toward serious action. 

Platforms should commit to reporting on the progress made on battling disinformation and hate, with publicly and with independently verifiable metrics. 

Informed customers are protected customers. Is that Twitter account a bot? Is that trend engineered? These are not as hard to figure out as many still think. Newish tools make it relatively easy to flag automated and other inauthentic attempts to steer conversation and news online. 

That information is not, however, being provided to users in an easy manner. Just as platforms tag accounts with all sorts of metrics of everything from geographic location to interests, platforms should notify users of the degree of automation of all social media accounts. A marker on every social media account that presents how much of each account’s activity is automated would thus clarify whether accounts more likely belong to authentic persons or organizations as opposed to bots, trolls, or sock puppets. The same could be marked on “trending” topics, allowing users to see how much of the trend was authentic and how much of it is just engineered engagement. 

Empowered customers are protected customers. Platforms could also allow users to select filters for the content they see. For instance, users should have an option to filter out trending online conversations that receive heavy participation from automation and inauthentic networks. This will help social media users to engage with what is real and avoid what is the creation of automation, which would also help alter the system that rewards manipulation. 

These three measures certainly won’t end the problem, for the very reason that there are no silver bullets in any battle. But we can easily introduce into the battle against hate and falsehood online more of what the German philosopher of battle Carl von Clausewitz called friction: “Everything in war is very simple, but the simplest thing is difficult.” The opposite has so far been true for those pushing disinformation. 

Let’s make it more difficult for toxic actors online and simpler to identify and push back against them. 

Welton Chang is Chief Technology Officer at Human Rights First.

Doowan Lee is an adjunct professor at the University of San Francisco and senior director of research and strategy in a Silicon Valley firm. 

P.W. Singer is Strategist at New America and co-author of LikeWar: The Weaponization of Social Media