The Tuskegee of the Internet: Facebook's Undisclosed Research on Teen Users

In a slide deck prepared for internal use in 2019, a team of Facebook researchers summed up their findings about the company’s own product in a single, damning line: “We make body image issues worse for one in three teen girls.” The slides were never meant to leave the building. They surfaced in September 2021, when a former Facebook product manager named Frances Haugen walked out with tens of thousands of pages of internal documents and handed them to the Wall Street Journal, the US Securities and Exchange Commission, and eventually a congressional subcommittee. In the days that followed, a phrase began circulating among critics and commentators: this, they said, was “the Tuskegee of the internet”. It is a heavy comparison, and it is worth taking seriously in both directions — for what it gets right about a real, documented wrong, and for the ways it strains against the history it invokes.

What Facebook actually knew, and when

Set the analogy aside for a moment and look at the record, because the record is solid. Facebook, which had by then renamed its parent company Meta, conducted extensive internal research into how Instagram affected the mental health of teenagers, particularly girls. This was not speculation by outsiders. It was the company studying its own users, with its own staff, and reaching conclusions that its own leadership then chose not to disclose.

The documents Haugen released, published by the Journal as “The Facebook Files” beginning in September 2021, contained a series of internal presentations from 2019 and 2020. One reported that among teenagers who felt suicidal, a meaningful share traced the desire back to Instagram. Another found that 32 per cent of teen girls said that when they felt bad about their bodies, Instagram made them feel worse. A third noted that the app’s harms were specific and hard to design away, because the very features that drove engagement — the endless scroll of curated, filtered, idealised images — were the same features causing the damage. Internal researchers wrote candidly that “comparisons on Instagram can change how young women view and describe themselves”.

The confirmed wrongdoing, then, is a matter of disclosure and testimony. In the spring of 2021, before the leak, a Meta executive had told a US congressional hearing that the company’s research showed using its apps could have positive mental-health benefits, and had characterised the effects on teenagers in reassuring terms. The internal documents told a considerably darker story than the public one. When the SEC complaint Haugen filed alleged that Meta had misled investors and the public about what it knew, it was pointing at exactly this gap — between the calming account offered to legislators and the alarming account circulating on the company’s own servers.

There was more. A separate strand of the story, reported by The Australian back in 2017, described a leaked internal Facebook document in which the company told advertisers it could identify moments when teenagers felt “worthless”, “insecure”, “stressed”, “defeated”, “anxious” and “like a failure” — the implication being that emotionally vulnerable young people could be targeted at their lowest ebb. Facebook said at the time that it did not offer tools to target people based on emotional state and that the research should not have been done. Whether or not the targeting product existed as described, the document confirmed that the company was studying the emotional vulnerability of minors and discussing it with commercial partners.

The fork: why “Tuskegee” both fits and overreaches

The comparison to the Tuskegee syphilis study is doing real work, and it is worth being precise about which part holds and which part breaks, because the imprecision is where the mythology creeps in.

What the analogy captures is genuine. In both cases an institution possessed knowledge of harm to a vulnerable population, and in both cases it chose concealment over disclosure. The US Public Health Service ran its study at Tuskegee from 1932 to 1972, watching syphilis progress in hundreds of Black men who were never told they had a treatable disease, long after penicillin became the standard cure. Facebook had internal evidence of harm to teenagers and, rather than publish it or act decisively on it, kept it in slide decks and told the public a rosier version. The shared structure — knowledge, vulnerability, silence — is what makes people reach for the parallel, and it is not an empty reach.

But the differences are not trivial, and honesty requires naming them. Tuskegee was a deliberate research protocol whose design required withholding treatment; the harm was the intended mechanism of the study, and it killed people, causing dozens of deaths and infecting wives and children over four decades. Facebook did not design Instagram in order to injure teenagers, and its researchers were, if anything, the ones sounding the alarm internally. The harm at issue is diffuse, correlational, and genuinely contested in its magnitude — a matter of body-image distress and self-reported wellbeing, which sits on a very different plane from an untreated infectious disease that ends in death. The scientific debate about how much social media actually damages adolescent mental health, as opposed to how much it reflects or amplifies distress that has other causes, remains live and unsettled among researchers who have no stake in defending Meta.

So the fork is this. The documented core — a company that knew its product hurt some young users and did not level with the public — is real and serious. The mythologised version, in which “Tuskegee of the internet” implies a deliberate, lethal experiment consciously inflicted on children, imports a moral weight the facts do not carry. The analogy illuminates the shape of the failure while overstating its kind. That distinction is not pedantry. It is the difference between holding a company accountable for what it actually did and reaching for the worst historical crime available because the real one, though bad, is harder to make vivid.

The journey: from a slide deck to a Senate hearing

The path this story travelled says something about how corporate secrets break in the platform era. Nothing here was uncovered by outside investigators piecing together fragments. The knowledge existed, fully formed, inside the company, written down by the company’s own people. It became public because one employee decided the gap between the internal truth and the external message was intolerable and carried the evidence out of the building.

Frances Haugen’s testimony to the US Senate Commerce Committee on 5 October 2021 was measured and, for that reason, devastating. She did not describe a cartoon villain. She described a company that, she said, consistently resolved conflicts between its own profits and the safety of its users in favour of profits — a choice she attributed to the ordinary gravity of incentives rather than to any cartoon malice. That framing — competent people inside a well-run organisation making individually defensible choices that add up to a documented harm — is the same pattern that runs through nearly every confirmed corporate cover-up. It is the pattern of the tobacco industry’s own research, where scientists inside the companies understood the danger years before the public did, and the machinery of the firm quietly converted that knowledge into public denial.

The response from Washington told its own story about how these disclosures land. Within days of Haugen’s testimony, senators from both American parties — a rare alignment — were competing to denounce the company, and a subcommittee that had struggled for years to agree on anything regarding technology regulation suddenly found common ground. Meta’s own head of global safety, Antigone Davis, had testified to the same body just days earlier and been met with open scepticism. Internationally, the documents fed directly into the drafting of online-safety legislation, including provisions in the United Kingdom’s Online Safety Bill aimed specifically at protecting children from algorithmic harm. Whatever one concludes about the precise magnitude of Instagram’s effect on teenagers, the leak demonstrably moved the machinery of law, which is a measure of how seriously legislators took the gap between what the company knew and what it had said.

The “Tuskegee of the internet” label spread because it gave the story a handle. A slide reading “one in three teen girls” is a statistic; a comparison to one of the most notorious ethics violations in American history is a headline. The phrase let commentators express the moral seriousness they felt about the disclosure, and it travelled faster than the careful caveats. That is how these things move: the vivid analogy outruns the precise account, and within weeks the shorthand has replaced the substance in most people’s memory of what happened.

What it was really about

Underneath the affair sits an anxiety that any parent of a teenager will recognise viscerally — the sense that a device in a child’s hand is doing something to them that the adults around them cannot see, measure, or switch off, and that the people who can see it have reasons not to say. The documents gave that formless worry a hard edge. Here was proof that the company itself had measured the thing parents feared, found it real for a meaningful fraction of kids, and kept the measurement quiet while its executives offered reassurance.

That is why the Tuskegee comparison resonated even where it did not quite fit. Tuskegee is, in American memory, the archetype of the trusted institution that knew and stayed silent while the vulnerable were harmed — the reason a great many people carry a rational, historically earned distrust of official reassurance. Reaching for it was a way of saying: we were told to trust you with our children, and you knew something you did not tell us. The precision of the analogy mattered less than the emotional truth it was trying to name, which was a betrayal of trust by an institution that had asked for that trust and traded on it.

The most useful thing to hold from the episode is the specific, provable fact underneath the slogan, once the slogan has been set aside: a company studied vulnerable young people, learned its product was hurting a portion of them, and chose to keep that knowledge in-house while presenting a gentler story in public. That is enough. It does not need to be Tuskegee to be a genuine wrong, and insisting on the loudest possible historical parallel risks making the accountability depend on an exaggeration that the company’s defenders can then knock down. The buried research was real. The teenagers were real. The silence was a choice, made by people who had the numbers in front of them and decided the public did not need to see them. Those facts stand on their own, and they are the ones worth remembering when the analogy has faded and the slogan has done its work.