How Facebook’s algorithms promote hate and drive toxic content

Facebook has been in the limelight for two issues of late—both damaging from the company's perspective, but in terms of public interest, each has its own level of usefulness. The news item with less long-term significance but more sensational media appeal is that what was supposed to be a small configuration change took Facebook, Instagram and WhatsApp down for a few hours on October 4. It affected billions of users, showing the world how important Facebook and other tech giants have become to many people's daily lives and even to the operation of small businesses. Of course, the much more significant news is the Facebook whistleblower Frances Haugen, a former employee of the company, who made tens of thousands of pages of Facebook's internal documents public. These documents showed that Facebook's leadership repeatedly prioritized profits over social good. Facebook's algorithms polarized society and promoted hate and fake news because they drove up "engagement" on its platforms. That the platform is tearing apart communities, and even endangering teens, especially girls, for not having "perfect" bodies, apparently mattered not a jot to Facebook.

The Wall Street Journal has published detailed exposés quoting Facebook's internal documents and Frances Haugen, who has also appeared on CBS' "60 Minutes" and in congressional hearings. "The thing I saw at Facebook over and over again was there were conflicts of interest between what was good for the public and what was good for Facebook," Haugen told CBS correspondent Scott Pelley on "60 Minutes." "And Facebook, over and over again, chose to optimize for its own interests, like making more money."

The 37-year-old data scientist has filed eight whistleblower complaints against Facebook with the Securities and Exchange Commission (SEC) with the help of a nonprofit organization, Whistleblower Aid. These complaints are backed by hard evidence: tens of thousands of internal Facebook documents Haugen had secretly copied before leaving Facebook.

Why is this big news when these issues relating to Facebook have been raised time and again, and were more prominently highlighted after revelations regarding the data firm Cambridge Analytica and Facebook became public in 2018? Did we not already know how Facebook, WhatsApp and other social media platforms have become powerful instruments today that help promote hatred and divisive politics? Have the UN investigators not held Facebook responsible for the genocidal violence against Rohingyas in Myanmar? Were similar patterns not visible during the communal riots in Muzaffarnagar, in the Indian state of Uttar Pradesh in 2013 and 2017?

The big news is that we now have evidence that Facebook was fully aware of what its platform was doing. We have it from the horse's mouth: internal Facebook documents that Haugen has made public.

By prioritizing posts that promote "engagement"—meaning people reading, liking or replying to posts on Facebook, WhatsApp and Instagram—Facebook ensured that people stayed on its platform for much longer. Facebook users could then be "sold" to the advertisers more effectively, by showing them more ads. Facebook's business model is not promoting news, friendly chitchat among users, or entertaining people. It is selling its users to those who can sell them merchandise. And like Google, it has a far better understanding of who its users are and what they may buy. This is what provided Facebook with 98 percent of its revenue in 2020 and has made it one of the six trillion-dollar companies (as of September 2021) in terms of market capitalization.

Testifying before Congress on October 5, Haugen said that "Facebook uses artificial intelligence to find dangerous content," Ars Technica reported. "The problem is that 'Facebook's own research says they cannot adequately identify dangerous content. And as a result, those dangerous algorithms that they admit are picking up the extreme sentiments, the division[s].'"

That this was happening is widely known and has been discussed, including in my own columns. Facebook's response to this criticism was that they were setting an independent supervisory board for oversight and employing a large number of fact-checkers. This and other processes would help filter out hate posts and fake news. What they hid was that all these actions were simply cosmetic. The driver of traffic, or what a person sees in their feed—or, in Facebook's terms, what they engage with—is determined by algorithms. And these algorithms were geared to promote the most toxic and divisive posts, as this is what attracts engagement. Increasing engagement is the key driver of Facebook's algorithms and defeats any measure to detoxify its content.

Haugen's congressional testimony also highlights what the real problems with Facebook are and what governments around the world must do in order to protect their citizens: to make Facebook accountable, not by censoring hate speech and fact-checking misinformation posted by individual users, but rather by targeting their algorithms' tendency to enable the dangerous high-engagement content. "This is not simply a matter of certain social media users being angry or unstable, or about one side being radicalized against the other," she said. "These problems are solvable… Facebook can change, but is clearly not going to do so on its own." While addressing the U.S. Congress about what can be done to regulate Facebook nationally, Haugen also acknowledged the problems Facebook's algorithms have caused worldwide. The solution, therefore, must also be global. In her testimony, she said that Facebook's meager proposed self-reforms would be insufficient to making the company accountable for its actions until they are made fully transparent. Facebook is hiding behind "safe harbor" laws that protect tech companies like Facebook, who do not generate content themselves, but provide their platform for what is called user-generated content. In the U.S., it is Section 230 of the Communications Decency Act that allows these tech companies to "moderate content on their services"; in India, it is Section 79 of the Information Technology Act. Both countries are considering reforms.

In the U.S., "a Section 230 overhaul… would hold the social media giant responsible for its algorithms," Ars Technica reports. In Haugen's words, "If we had appropriate oversight, or if we reformed [Section] 230 to make Facebook responsible for the consequences of their intentional ranking decisions, I think they would get rid of engagement-based ranking.… Because it is causing teenagers to be exposed to more anorexia content, it is pulling families apart, and in places like Ethiopia, it's literally fanning ethnic violence." The key problem is not the hateful content users generate on Facebook; it is Facebook's algorithms that drive this poisonous content to a person's Facebook feed continuously to maximize the company's advertising revenue.

Haugen added:

"Facebook wants to trick you into thinking that privacy protections or changes to Section 230 alone will be sufficient. While important, these will not get to the core of the issue, which is that no one truly understands the destructive choices made by Facebook except Facebook. We can afford nothing less than full transparency. As long as Facebook is operating in the shadows, hiding its research from public scrutiny, it is unaccountable. Until the incentives change Facebook will not change. Left alone, Facebook will continue to make choices that go against the common good, our common good."

Of course, the widespread prevalence of toxic content on Facebook's platforms is helped by its willful neglect of not having language classifiers—the algorithms used to detect hate speech—for content that is not in English and is created in other languages. Even though Hindi is the third most spoken language in the world and Bengali is the sixth, according to Haugen, Facebook does not have enough "hate speech classifiers" in these two languages.

I have previously written why divisive content and fake news have more virality than any other content. Haugen's documents confirm what analysts including myself have been saying all along. The algorithms that Facebook and other digital tech companies use today do not directly code rules to drive up engagement. These companies instead use machine learning, or what is loosely called artificial intelligence, to create these rules. It is the objective—increasing engagement—that creates the rules that lead to the display of toxic content on the users' feeds that is tearing societies apart and damaging democracy. We now have hard evidence in the form of the leaked documents that this is indeed what has been happening. Even worse, the Facebook leadership and Mark Zuckerberg have been fully aware of the problem all along.

Not all the harm on Facebook's platform, however, was caused by algorithms. From Haugen's documents, we find that Facebook had "whitelisted" high-profile users whose content would be promoted even if they violated Facebook guidelines. Millions of these special users could violate Facebook's rules with impunity. I had earlier written on evidence provided by the Wall Street Journal about how Facebook India protected BJP leaders in spite of repeated red flags relating to their posts being raised within Facebook itself.

This is not all that Haugen's treasure trove of Facebook's internal documents reveal. Reminiscent of cigarette companies research on how to hook children to smoking, Facebook had researched "tweens," who are children in the age group of 10 to 12. Their research was on how to hook the "pre-teens" to Facebook's platforms so that they could create new consumers for its platforms. This is despite their internal research showing that Facebook's platforms promoted anorexia and other eating disorders, depression, and suicidal tendencies among teens.

All these facts should damage Facebook's image. But it is a trillion-dollar company and one of the biggest in the world. Its fat cash balance, coupled with the power it wields in politics and its ability to "hack" elections, provides the protection that big capital receives under capitalism. The cardinal sin that big capital may not tolerate is lying to other capitalists. The internal documents that Haugen has submitted to the SEC could finally result in pushback against social media giants and lead to their regulation—if not strong regulation, at least some weak constraints on the algorithms that promote hate on these social media platforms.

A decade-old quote is at least as relevant now in light of these recent Facebook developments as it was when then 28-year-old Silicon Valley tech whiz Jeff Hammerbacher first said it: "The best minds of my generation are thinking about how to make people click ads." This has long been the beating drum driving the march of social media giants to their trillions.

This article was produced in partnership by Newsclick and Globetrotter. Prabir Purkayastha is the founding editor of Newsclick.in, a digital media platform. He is an activist for science and the free software movement.