I don’t often agree with the editorial board at the Washington Post, but I did last week. An editorial over the holiday weekend titled “Schools should ban smartphones. Parents should help” listed just a handful of the numerous problems an ever-present screen has caused for children and their ability to learn. 

It is practically impossible to prevent children from using phones in class when they are allowed to use them at every other point. The self-control muscle is weak in adults when it comes to screens; we should not be surprised that it is nonexistent in young children. (Ask yourself why it is the parents, not the children, who are most “enraged” at the suggestion that little Billy shouldn’t be allowed to check his notifications in the bathroom stall.) Banning phones in school completely, meanwhile, has been shown to make a world of difference in educational outcomes. 

A report out of the Wall Street Journal Monday gave a fresh edge to why caution is so necessary. After creating new accounts to test the Instagram reels algorithm, the Journal found the app promoted sexually explicit and pedophilic content to accounts that followed young cheerleaders, gymnasts, and teen and pre-teen influencers. If the test accounts followed some of the (mostly adult male) accounts that follow these pre-teen influencers, the app sent an even bigger flurry of lewd reels. With only a whiff of provocation, Instagram attempted to send its users—who, based on the criteria, could easily have been underage—down a dark and wicked spiral.  

Obviously, this test was designed to produce just such a response, though it was easily replicated by the Canadian Centre for Child Protection with similar results. Underage girls in cheerleading garb are promiscuous enough that you’d be concerned to find your husband following them, but innocuous enough to slide beneath the radar when mixed in with other accounts, and therefore a perfect test case of the algorithm’s inclination. Clearly, as the Journal demonstrated, that inclination is on the side of promoting vice. 

For those who have been keeping score, this is not Meta’s first pedophilic infraction. Back in June, the Journal reported that Meta algorithms on both Facebook and Instagram were connecting large communities of users that it suspected may be interested in pedophilic content. In response, Meta set up an internal task force, purportedly to detect more of this suspicious content and to suspend more harmful accounts. After this week’s discovery, however, Meta declined to comment to the Journal on why the algorithm still promotes underage sexual content. Instead, the social media conglomerate pointed to safety tools and the employees who remove or reduce the prominence of some 4 million harmful videos each month. They claimed the Journal’s tests “produced a manufactured experience.”

This is the typical refrain from social media companies under fire: How can Instagram be held accountable for every single piece of media posted on its site? How indeed, when it is given free speech protections in the case of bad content, yet the old private company garb when it comes to censorship of unpopular takes. You have heard this story countless times before. 

The words of Meta employees to the Journal, meanwhile, are chilling: Meta knows its Instagram algorithm promotes pedophilia, because the mechanism that does so is the same one that gets the rest of us otherwise-self-controlled adults addicted to scrolling. Employees told the Journal that “preventing the system from pushing noxious content to users interested in it…requires significant changes to the recommendation algorithms that also drive engagement for normal users.” 

In other words, promoting vice is not a bug, but a feature: This is how the engine was designed to work. And, as the advertisements interspersed between salacious reels in the Journal’s test suggest, the highway to hell is paved with corporate sponsorships. 

In this sense, expecting Meta to control just one area of its algorithm is not unlike expecting grade school children to not touch their phones during class, even though they have free access to them throughout the school day. The incentive structure is aligned against it. No internal task force will change that.

Image-based social media has always trended toward such voyeurism. The reason pre-teen influencers are flashing their midriffs for thousands of adult male followers in the first place is that this is what the algorithm has taught them to do. This is what gets views, and views equal money, sponsorships, and most importantly, respect among their peers. The writer Mary Harrington has summed up this tendency of social media to push users toward mental and physical nakedness as the “clout economy,” in which the way to success is paved with metaphorical “pornography of the self.” Clearly, that is not just a metaphor. 

What is needed, then, is a genuine reflection from our legislators on how such apps damage the American people in mind and in body, ripping apart our social fabric by promoting such vice. 

More simply, as the Heritage Foundation’s Big Tech Campaign Lead Wes Hodges put it to me, “we cannot trust social media to self-regulate.” 

“When repeatedly confronted with reports of social media companies failing to self-regulate to the level of safety expected in any other public space, we must align their incentives with the well-being of American families,” Hodges said. “This duty is doubly true when there is a profit motive behind child access to explicit content and pedophilic algorithms, as highlighted by the WSJ report on Instagram.”

This could mean several things, from raising the age of social media access to amending Section 230 to ensure criminal exploitation is not protected, according to Hodges. But the top line is not variable: Algorithms are not benign. They cannot be left to govern themselves.