de-amplify_
← all distillations

distillation · the evidence ledger

Congressional Hearing: Social Media and the Teen Mental Health Crisis (Nov 7, 2023)

Senate Judiciary Subcommittee on Privacy, Technology, and the Law; Nov 7, 2023; witness Arturo Bejar (former Facebook/Meta engineer); S.Hrg. 118-663

This is the raw, evidence-tiered claim ledger behind the hearings hub. It is published for transparency: the source of truth, with its tiers intact, current as of 2026-07-16.

how to read the tiers

Each claim is tagged by the strength of the evidence behind it, not by how certain it sounds.

  • [ESTABLISHED]: stated in a primary source (a court order or filing, the docket, an official release) or, for a hearing, said on the official record. A fact about what was filed or said, not necessarily adjudicated true.
  • [OBSERVED]: carried by secondary coverage, or by a document read into the record.
  • [ASSUMED]: the distiller's inference, flagged as such.
  • Provenance flags ([interested-party], [advocacy-witness], [single-witness], [lawmaker characterization]) mark a claim that rests on one interested voice: reported, not adjudicated.

Reader summary (derived, plain text)

On November 7, 2023, a Senate Judiciary subcommittee heard sworn testimony from Arturo Bejar, an engineer Facebook had hired twice to make its products safer for kids. His account, backed by his own memos to Mark Zuckerberg and Adam Mosseri that were entered into the record, was that Meta measured the harm children experienced, knew the numbers, and repeatedly chose not to act. His single most load-bearing claim for anyone studying platform safety: the safety features the company shipped were, in his words, "a safety feature in name only to placate the press and regulators." He described a help button he asked the company to build so a child could flag a hurtful message and get help, and it was not built. A striking detail: he proposed, in his words, "let's measure our help by whether it helped", and that was not adopted. Senators from both parties, Democrats and Republicans, agreed the problem is the product's design and the absence of any incentive to change it. One caution for the reader: this hearing had no witness from Meta or any platform, so the companies' side appears only as quoted documents, not live answers. Their live testimony is a separate hearing (the January 31, 2024 CEO hearing) and their legal position is in the lawsuits. As of the date above, this summary reflects the record of that one hearing.

Brake-integrity relevance (why this hearing feeds the de-amplify thesis)

This hearing is the clearest on-the-record statement, from inside the company, of the exact pattern the brake-integrity work is about: a control that looks like a brake but does not stop the car.

  • The fake brake: Bejar's "safety feature in name only". Tools shipped to answer public criticism, defined so narrowly they did not reduce the harm children actually reported.
  • The real brake, refused: the help button a child could press when a message hurt them. As he put it, "And it doesn't matter what the content is, that child deserves help." He asked for it. It was not built.
  • The self-test that failed: "let's measure our help by whether it helped" ... "and that was not adopted." A brake you never test against the thing it is supposed to stop is a sticker.

These map directly onto the movement's homepage self-test (set a limit, close the app, reopen, did it hold?). The hearing supplies the insider evidence that the answer was engineered to be "no."

Claims (the ledger)

What the hearing established (on the record)

  • [ESTABLISHED] The hearing was held Nov 7, 2023 before the Senate Judiciary Subcommittee on Privacy, Technology, and the Law, chaired by Sen. Richard Blumenthal (D-CT), ranking member Sen. Josh Hawley (R-MO). (Transcript header and opening.)
  • [ESTABLISHED] The sole witness was Arturo Bejar, former director of engineering for "Protect and Care" at Facebook (2009 to 2015), who returned as a consultant to Instagram's Well-Being Team (2019 to 2021). He testified under oath. (Blumenthal introduction; Bejar statement.)
  • [ESTABLISHED] Bejar testified that in a 2021 Instagram survey, one in eight children aged 13 to 15 reported an unwanted sexual advance on the platform in the last seven days. [single-witness, but drawn from an internal survey he ran; the memo was entered into the record] (Bejar statement; Hawley Q&A.)
  • [ESTABLISHED] Bejar testified he emailed the finding to Mark Zuckerberg, Sheryl Sandberg, Adam Mosseri and other executives on Oct 5, 2021, and that Zuckerberg did not reply and did not meet with him. (Recipients per Blumenthal's opening statement; the no-reply and no-meeting exchange per Hawley Q&A. The Oct 5, 2021 email was entered into the record.)
  • [ESTABLISHED] Bejar reached three conclusions on the record: that Meta knows the harm kids experience and that its measures fail to address it; that there are actionable steps Meta could take; and that the company decides "time and time again" not to take them. (Bejar statement.)
  • [ESTABLISHED] Sen. Blumenthal stated in his opening that when users reported harmful content to Facebook, it "took action only 2 percent of the time". [lawmaker characterization: the chair's assertion presenting the committee's evidence, not witness testimony] (Blumenthal opening statement.)

The mechanism (thesis-relevant, all Bejar testimony unless noted)

  • [ESTABLISHED] Bejar testified that safety features developed in response to public outcry were "in reality, kind of a placebo--a safety feature in name only to placate the press and regulators", built on "very deliberately narrow definitions of harm." [single-witness] (Bejar statement.)
  • [ESTABLISHED] He testified that the narrow definitions let the company grade itself: "The company was creating its own homework." [single-witness] (Bejar statement.)
  • [ESTABLISHED] He testified that most of the child-safety tools built during his earlier tenure "had been removed" by the time he returned in 2019. [single-witness] (Bejar statement.)
  • [ESTABLISHED] He testified that the internal term for addiction-like use was "problematic usage", and that its definition was too narrow to capture what parents observe. [single-witness] (Coons Q&A.)
  • [ESTABLISHED] He testified he proposed the team "measure our help by whether it helped", and that it "was not adopted." [single-witness] (Coons Q&A.)
  • [ESTABLISHED] He testified he repeatedly asked the company to add a help button a child could press on a harmful direct message, independent of content, and it was not built. [single-witness] (Welch Q&A; Butler Q&A; Hawley second-round Q&A.)
  • [ESTABLISHED] He testified that the user-feedback brake exists for advertisers but not for children's direct messages: "You can take an ad and say that is sexually inappropriate. But there's no way for a child to do that when they get a message". The reporting mechanism the safety systems depend on is withheld exactly where kids are harmed. [single-witness] (Hawley Q&A; restated in Butler Q&A.)
  • [ESTABLISHED] Asked whether some forms of social media "optimize for engagement", he agreed, adding in the senator's own term that "they reward being an a-hole." (Kennedy Q&A.)
  • [ESTABLISHED] He testified his daughter kept a harassing interaction up because reporting would do nothing and deleting it would cost her reach: "I will not delete it because I'm worried that that will mean that less people will see my posts." The product's incentives punish self-protection. [single-witness] (Kennedy Q&A.)
  • [ESTABLISHED] He located the company's metric choice at the top: "prioritizing prevalence over harm, is something that Mark sets direction for that whole executive team." This is the internal counterpart of "creating its own homework." [single-witness] (Blackburn Q&A.)
  • [ESTABLISHED] On discovering that chief product officer Chris Cox already knew the harm numbers: "The first person to do that was Chris Cox. And I found it heartbreaking because it meant that they knew and that they were not acting on it." [single-witness] (Blumenthal Q&A.)
  • [ESTABLISHED] He testified that "there are no goals to reduce unwanted sexual advances, as far as I am aware." [single-witness] (Hirono Q&A.)
  • [ESTABLISHED] He testified the fix was not a matter of significant investment: "It is a matter of how much they prioritize the work, and whether they're willing to set their goals based on what teens are experiencing." [single-witness] (Hawley second-round Q&A.)
  • [ESTABLISHED] Asked by Sen. Graham whether it is fair to say that "in its current form what you're describing is a dangerous product", Bejar answered "Correct." (Graham Q&A.)
  • [OBSERVED] The record includes a Wall Street Journal finding (June 2023), read aloud by Sen. Hawley, that "Instagram's algorithms promote" networks devoted to underage sexual content, not merely host them. (Hawley Q&A. This is outside reporting cited at the hearing, not independent verification by the committee.)

The platforms' own words, entered into the record (claim vs later reality)

  • [OBSERVED] Adam Mosseri (then head of Instagram) told a Senate subcommittee in Dec 2021 that Instagram had "built tools that prevent bullying from happening in the first place, and empower people to manage their accounts so they never have to see it" (his prior testimony, read into this hearing's record by Sen. Blackburn).
  • [ESTABLISHED] Bejar, who had consulted for Instagram's Well-Being Team, called that characterization "profoundly misleading" on the record. (Blackburn Q&A.)
  • [OBSERVED] Mark Zuckerberg wrote in an Oct 2021 note (quoted from the Massachusetts complaint entered into the record) that "everything we build is safe and good for kids". The same note said that "when it comes to young people's health or well-being, every negative experience matters." Bejar, holding up that note, responded: "And I believe that is what Instagram does today." He meant the harm, not the care. (Blumenthal second-round Q&A.)

Quotes (attributed record)

Every quote is a verified verbatim substring of the official transcript (S.Hrg. 118-663). Speaker, role, the moment in the hearing, and the line range in the cached transcript (sources/CHRG-118shrg60432.txt) are given so each can be located and re-verified.

The witness: Arturo Bejar (former Facebook/Meta safety engineer)

  • "a safety feature in name only to placate the press and regulators" [mechanism: the fake brake, named from inside the company] (opening statement; transcript lines 622-624)
  • "The company was creating its own homework." [mechanism: narrow harm definitions let the company grade itself] (opening statement; transcript line 627)
  • "Meta knows the harm that kids experience on their platform, and the executives know that their measures fail to address it." [mechanism: knowledge, not ignorance] (opening statement; transcript lines 640-642)
  • "they are deciding time and time again to not tackle these issues" [mechanism: a choice, not a limitation] (opening statement; transcript lines 645-646)
  • "let's measure our help by whether it helped" ... "and that was not adopted." [mechanism: the safety tool was never tested against whether it worked] (Q&A, Sen. Coons D-DE; transcript lines 1820-1821)
  • "can we please add a button when a child receives this message that says like, ``Please help me.''" [mechanism: the real brake he asked for and did not get] (Q&A, Sen. Welch D-VT; transcript lines 1538-1539)
  • "Those are not a matter of significant investment. It would not cost them as much. It is a matter of how much they prioritize the work" [mechanism: the fix was cheap; the refusal was a priority choice] (second-round Q&A, Sen. Hawley R-MO; transcript lines 2045-2047)
  • "It's a product. It needs to be different. It has to change." [mechanism: frames the issue as product design, not speech] (second-round Q&A, Sen. Blumenthal D-CT; transcript lines 1951-1952)
  • "I agree with that they make research. I don't agree that they make changes." [mechanism: research without action] (Q&A, Sen. Blackburn R-TN; transcript lines 1367-1368)
  • "algorithms are as good as their inputs" [mechanism: an automated safety system blind to what kids flag cannot catch it] (Q&A, Sen. Hawley R-MO; transcript line 856)
  • "You can take an ad and say that is sexually inappropriate. But there's no way for a child to do that when they get a message" [mechanism: the feedback brake exists for advertisers and is withheld from children's DMs; the most concrete receipt in the hearing] (Q&A, Sen. Hawley R-MO; transcript lines 860-862)
  • "The first person to do that was Chris Cox. And I found it heartbreaking because it meant that they knew and that they were not acting on it." [mechanism: knew-and-did-not-act, tied to a named executive] (Q&A, Sen. Blumenthal D-CT; transcript lines 742-744)
  • "prioritizing prevalence over harm, is something that Mark sets direction for that whole executive team." [mechanism: the metric choice that produces the fake brake, set at the top] (Q&A, Sen. Blackburn R-TN; transcript lines 1425-1426)
  • "I will not delete it because I'm worried that that will mean that less people will see my posts." (Bejar recounting his daughter) [mechanism: reach incentives punish self-protection] (Q&A, Sen. Kennedy R-LA; transcript lines 1209-1211)

Lawmakers naming the mechanism (bipartisan, not the theater lines)

  • "the Kids Online Safety Act is also about the product. It's about product design." Sen. Richard Blumenthal (D-CT), second-round Q&A (transcript lines 1937-1938). [frames the remedy as design, not censorship]
  • "All six passed unanimously. Every Democrat and every Republican." Sen. Richard Durbin (D-IL), opening statement (transcript lines 443-444), on the child-safety bills the Judiciary Committee had already reported. [the problem is treated as bipartisan on the record]
  • "that statistic is designed to mislead" Sen. Josh Hawley (R-MO), opening statement (transcript lines 397-398), on the company's public claim to take down roughly 90 percent of harmful material. [names the safety metric itself as the misleading artifact]
  • "They had built tools. You built them. That was true. But they chose to remove that." Sen. Marsha Blackburn (R-TN), second-round Q&A (transcript lines 2250-2252). [the tools existed and were withdrawn: a choice]

Tensions / open questions

  • One witness, one side. Meta and every other platform were absent. The strongest mechanism claims (placebo features, removed tools, ignored memos) are single-witness. They are corroborated in part by documents Bejar authored and by the WSJ reporting, but the company had no chance to answer them here. The honest posture is "Bejar testified that," not "it is established that Meta did."
  • The narrow-definition-of-harm knife cuts both ways. Sen. Hirono (D-HI) noted on the record that Meta has been accused of over-censoring some content (she cited health advertisements related to menstruation) even as it under-acts on youth harm. This complicates a simple "they never moderate" story: the claim is not that the company cannot act, but that it aims its action narrowly. Worth holding, because it is the fairest read and the more precise one.
  • Remedy is contested even among allies. Senators pushed several different levers (KOSA product-design rules, Section 230 repeal, a private right of action, transparency mandates). Bejar repeatedly declined to endorse a specific legal remedy ("I'm not a lawyer"), staying on the design and transparency ground he could speak to. The hearing establishes the problem more than it settles the fix.
  • Time-sensitive political claims. Several senators pledged floor votes "before the end of the year" (2023). Those pledges are on the record; whether they were kept is outside this hearing and should not be read from it.

Sources

← the hearings hub

all evidence ledgers