I’ve spent years working at some of the world’s largest social media platforms. I’ve seen how the right kind of moderation can improve conversations, and how the wrong kind — or none at all — can produce chaos, spread hate and spill into real-world violence.
When I started in this field, the trust and safety teams I led had a simple premise: People should be able to express themselves so long as they do not put others in danger. Over time, though, I came to appreciate the complexity of defining danger in a digital environment with billions of users.
The difficulty of this work was on stark display when the rapper formerly known as Kanye West, who now goes by Ye, posted a stream of misogynistic and antisemitic invective to his more than 30 million followers on the platform formerly known as Twitter, which now goes by X. In a dayslong spree recently, he posted uncensored clips from pornographic movies, denigrated women and Jews, posted photos of swastikas, named specific Jewish executives who he believed were sabotaging his career and said, “Slavery was a choice.”
My search of archived versions of Ye’s now-deleted posts suggests that only the post naming specific executives was flagged with a warning by X for potentially violating its hate-speech policies. Vilifying Jews individually and collectively and deploying the symbol of the regime that murdered six million Jews — using swastikas is expressly cited by X as an example of hateful imagery — seems to have gone unaddressed. That’s typical of what we’ve seen at Elon Musk’s X, which has shifted away from proactive enforcement of policies against hate, meaning sometimes intervening before content even becomes public or reducing the number of people who see it, and toward Community Notes, a crowdsourced system that empowers certain X users by adding context in the form of a label after the fact.
Meta, which owns Facebook, has also recently scaled back its interventions and is ending its third-party fact-checking program in the United States. Some cheer these moves as a return to freer speech. But what remains is the question of whether Ye’s speech should be broadcast via any platform’s algorithm. Where is the line between the right to speech and the right to reach?
I’ll be the first to acknowledge that no moderation system is perfect. I’ve sat in rooms where we have debated where to draw the line, knowing that to catch most harmful content we would also inadvertently remove innocent posts. Moderators can be unsure whether a satirical post crosses the line into hate speech or if a post expressing earnest concern about vaccine efficacy has veered into misinformation. There is no universal consensus on what constitutes “harm,” and without careful calibration of policies and the machine-learning models trained to enforce them, mistakes happen.
But we cannot let the perfect be the enemy of the good. Over the years, I’ve seen how hands-off approaches can give oxygen to hate speech and misinformation. When entire nations have faced political upheaval or public health crises, disinformation amplified on social media has directly compromised safety. Human and automated moderators, third-party fact-checkers and swift enforcement of clearly articulated policies can curtail the spread of harmful content. Failing to do so can contribute to genocidal rhetoric, such as the misinformation that fueled the genocide in Myanmar against the Rohingya.
Community-based moderation tools, such as X’s Community Notes or Reddit’s system of volunteer moderators, are valuable, but they can only supplement, not replace, proactive moderation by the platforms themselves. Millions of people saw Ye’s posts. If a platform relies heavily on community-driven moderation, and not other tactics that limit distribution or provide warnings, the damage is often done before anyone can intervene, as if you’d planned to make a hand-painted sign to stop a speeding train.
Relying on users to flag or fact-check content shifts the onus onto victims of harassment or misinformation. One of the most draining aspects of community moderation is that it often asks the very people in the line of fire to police the content that targets them. As I’ve seen firsthand, this can push people off platforms altogether, an exodus that weakens the diversity of voices required for a healthy civic discourse.
When social media first became mainstream, many dismissed it as a playground for personal photos and status updates. Today, it’s a communication hub where politicians campaign, businesses market and journalists break news. Without professional moderation, it’s too easy for toxicity to flourish, for people with intent to harm to take advantage and for foreign bots to hijack the national conversation. Even deleted content lingers, retweeted and screenshot, fueling bigotry that can embolden others. Community Notes might eventually offer context, but context isn’t always enough to quell the harm done.
As users, we, too, must be vigilant. We should report content that crosses the line, scrutinize sources before sharing dubious claims and support policies that uphold the free exchange of ideas without enabling abuse. But, just as we expect a city to have traffic lights, fire departments and emergency services, we should expect and demand that online environments are similarly protected.
Companies must invest in professionals who understand cultural context, language nuances and how threats evolve online. They should leverage emerging advanced A.I. systems that can examine text, images and other forms of communication, and also the context in which they are shared, to more accurately and consistently identify dangerous content and behavior. They should invest in getting this right, rather than scaling down moderation to cut costs or acquiesce to a particular political movement. And regulators or independent oversight bodies need the power and expertise to ensure these platforms live up to their responsibilities.
This isn’t about nostalgic longing for the old days of moderation; it’s about learning from failures and building a system that’s transparent, adaptive and fair. Whether we like it or not, social media is the public square of the 21st century. If we allow it to devolve into a battlefield of unchecked vitriol and deception, first the most vulnerable among us will pay the price, and then we all will.
Free speech is essential for a healthy democracy. But social media platforms don’t merely host speech — they also make decisions about what speech to broadcast and how widely. Content moderation, as flawed as it has been, offers a framework for preventing the loudest or most hateful from overshadowing everyone else.
The post Ye and the Limits of Free Speech Online appeared first on New York Times.
