Youth suicide, corporate scapegoating, and the legal shortcut that mistakes a defendant for a cause.
A young person’s suicide is a tragedy. It is not, by itself, proof against a technology company.
A screenshot is not a psychiatric history. A sequence of events is not necessarily a chain of causation. And a corporation’s bank balance is not medical evidence.
When a lawsuit elevates the last app a child used above the child’s developmental history, psychiatric symptoms, previous suicidal behavior, family vulnerability, and immediate personal crises, it risks doing something profoundly wrong: replacing an investigation with a target.
The question is not whether technology companies deserve sympathy. The question is whether allegations deserve to be treated as established facts.
Companies should answer for harm their conduct demonstrably caused. But neither public anger nor courtroom emotion should excuse an incomplete explanation of a death.
Start with the deaths—not the headlines
A major new review provides a useful corrective to simplistic explanations.
Published online September 4, 2026, in the Journal of the American Academy of Child & Adolescent Psychiatry, the meta-analysis by Daniel Dobin and colleagues pooled 19 psychological-autopsy study samples: 958 people who died by suicide and 1,102 community controls, all aged 25 or younger. Unlike research measuring general unhappiness or time online, these studies examined suicide deaths. (PubMed)
The five strongest associations reported in the abstract were:
| Factor | Pooled odds ratio | 95% confidence interval |
|---|---|---|
| Suicidal ideation | 16.1 | 5.0–51.5 |
| Previous suicide attempt | 14.2 | 6.3–31.8 |
| Acute stressful life events | 13.7 | 9.6–19.5 |
| Mood disorders | 9.9 | 5.3–18.2 |
| Substance abuse, using the authors’ terminology | 8.9 | 6.1–12.9 |
These figures come from the published abstract. They are associations, not percentages of responsibility, and the abstract does not establish that the estimates represent mutually independent effects. (PubMed)
The findings demand serious attention to what was happening in the young person’s life before death.
What psychiatric symptoms were present? Had there been a previous attempt? Was there an acute crisis? What changed in the preceding days or weeks? What do contemporaneous records show, rather than recollections assembled after litigation began?
Those questions are not distractions from a technology claim. They are essential to evaluating it.
Psychological-autopsy evidence also deserves scrutiny: retrospective reconstruction can be incomplete, and a pooled association does not resolve an individual case. But that limitation applies to the entire investigation. It is not permission to substitute an equally retrospective, selectively edited account of screen use.
An association is a reason to investigate. It is not a license to stop investigating.
The child had a history before the app
Genetic and familial vulnerability cannot be dismissed because it is inconvenient to a platform-centered theory.
A Swedish study involving approximately 2.76 million offspring found that shared genetic factors explained much of the intergenerational association in suicidal behavior, alongside parental behavioral-health problems and environmental factors. The study examined suicidal behavior, including attempts and deaths; it was not a platform study or a means of calculating responsibility for an individual death. (Nature)
The implication is straightforward: the psychiatric history does not begin when the account was opened.
A sound investigation should examine relevant family history, the onset of symptoms, earlier crises, treatment, and functioning. It should ask whether vulnerability and suicidal behavior were documented before the disputed exposure.
This is not genetic fatalism. Heritability is not destiny, and inherited vulnerability does not establish that a death was unavoidable. Nor does identifying a predisposition automatically establish that another exposure made no contribution. Population genetic estimates cannot be converted into percentages of legal fault. (Nature)
But the opposite distortion is equally unacceptable: treating a young person as a clinical blank slate until a video, recommendation, or chatbot appeared.
A complete explanation must account for the person who encountered the product—not merely the product the person encountered.
Relationships are not background noise
Family conflict, peer rejection, romantic loss, bullying, humiliation, and other interpersonal events deserve substantive investigation.
In Sheftall and colleagues’ study of suicide deaths among children and early adolescents in 17 states during 2003–2012, relationship problems with family members or friends were recorded in 60.3% of childhood cases and 46.0% of early-adolescent cases with relevant circumstances available. These were descriptive findings, not proof that relationships caused those percentages of deaths. Nevertheless, they show why relationships cannot be relegated to a footnote. (PubMed Central (PMC))
Dobin’s newer review strengthens attention to acute stress, although its category of “acute stressful life events” is broader than relationship problems and should not be relabeled without examining the underlying definitions. (PubMed)
The relevant investigation is concrete. What happened? When? How did the young person respond? Did the crisis precede increased online activity? Was the platform a source of the distress, a means through which the distress unfolded, or a place where the young person disclosed distress that already existed?
These possibilities must be tested, not assumed.
Examining relationships is not an accusation against bereaved parents. It is not an invitation to invent family dysfunction or conduct humiliating, irrelevant discovery. It is a demand for the same completeness that would be expected in any serious clinical assessment.
The answer to scapegoating a company is not scapegoating a family. The answer is examining the evidence.
ADHD, impulsivity, and anxiety require actual clinical analysis
ADHD deserves particular attention in younger children.
In the same Pediatrics study, among decedents with recognized mental-health problems, 59.3% of those aged 5–11 had recorded ADD/ADHD, compared with 29.0% of those aged 12–14. Depression showed the reverse pattern: 33.3% versus 65.6%. These figures describe diagnostic patterns within a subgroup, not the proportion of all deaths caused by each disorder. They demonstrate why a childhood-suicide analysis cannot simply borrow an adult diagnostic template. (PubMed Central (PMC))
Impulsivity also needs precision. Rash action during intense negative emotion is not identical to distractibility or sensation-seeking. A longitudinal study of more than 10,000 young adolescents associated negative urgency—acting impulsively under distress—and lack of perseverance with subsequent suicidal thoughts and attempts. It did not establish that all impulsivity measures were equivalent or explain suicide deaths. (PubMed Central (PMC))
Anxiety belongs in the assessment, but it should not be exaggerated either. A longitudinal meta-analysis found modest associations with later suicidal thoughts and attempts, but no statistically significant pooled association with suicide death. That finding concerned anxiety as measured in those studies, not proof that acute anxiety is irrelevant to a particular crisis. (PubMed)
Nor can depression and substance misuse be assigned exclusively to adult suicide. Dobin’s review found strong youth associations for both and reported stronger substance- and alcohol-abuse associations than in its adult comparison meta-analysis. (PubMed)
The organizing model is multifactorial. Replacing “the app caused it” with “ADHD caused it” would repeat the same mistake with a different label.
Screen time is not a diagnosis
The scientific problem with blanket accusations against social media begins with a basic distinction: ordinary use is not the same exposure as harassment, compulsive use, or a specific harmful interaction.
A 2024 JAMA Pediatrics meta-analysis covering 143 studies and more than one million adolescents found small average associations between social-media use and internalizing symptoms such as anxiety and depression. Those findings concern symptom associations—not the number of suicide deaths caused by a platform. (JAMA Network)
Nesi and colleagues’ 2021 meta-analysis found no statistically significant associations between general use frequency and the available outcomes of suicidal ideation, suicide plans, or nonsuicidal self-injury. The frequency evidence was limited, and no included frequency studies examined suicide attempts. In contrast, specific experiences—including cybervictimization and problematic use—showed associations with self-injurious thoughts and behaviors. (PubMed Central (PMC))
A 2025 JAMA study of 4,285 young people reached another important distinction: total screen time at baseline was not associated with the later outcomes studied, whereas high or increasing addictive-use patterns were associated with suicidal thoughts and behaviors. This was observational research, not a demonstration that ordinary viewing caused suicide deaths. Its authors also acknowledged that the analyses did not include several potentially relevant factors, including bullying, adverse childhood experiences, sleep disturbance, and stress. (DOI)
That is a very different account from “children use screens, children become distressed, therefore screens explain the deaths.”
There is contrary evidence that must be taken seriously. Another 2025 longitudinal study, involving 11,876 children and adolescents, found that increases in social-media use preceded greater depressive symptoms one year later. That strengthens concern about possible effects on mood. It does not establish that a particular platform caused a particular suicide. (DOI)
The defensible conclusion is not universal harmlessness. It is that a tally of viewing hours is a weak substitute for a clinically grounded account of what happened.
A small average association does not impose a ceiling on the harm of a particular interaction. Conversely, a harmful interaction does not establish that ordinary viewing is generally a major cause of suicide.
Nor can an odds ratio for acute stress be divided by a correlation for screen use to announce that one is a specified number of times more important. The studies measure different exposures, populations, and outcomes.
Bad arithmetic does not become good science because it favors the defense.
AI requires its own investigation
A companion chatbot, a general-purpose assistant, and a video feed are not interchangeable.
Research testing 25 consumer chatbots against simulated adolescent emergencies found substantial safety shortcomings, with companion chatbots performing worse than general assistants on several measures. That is evidence of deficiencies under the tested conditions. It is not a measurement of real-world suicide deaths attributable to AI. (JAMA Network)
A responsible analysis therefore cannot declare AI “proven trivial” by borrowing screen-time research. But it also cannot treat a disturbing excerpt as a complete explanation of a death.
The investigation should examine the authenticated conversation, its chronology, the historical product version, relevant safety responses, the user’s psychiatric course, and the alleged mechanism of harm.
Did the interaction intensify danger? Did it discourage help-seeking? Did it merely reveal an existing crisis? Those are questions for evidence—not conclusions to be selected according to which party retained the expert.
The law must demand proof—not a villain
The legal system already has a framework for scrutinizing expert claims.
Federal Rule of Evidence 702 requires sufficient facts or data, reliable methods, and reliable application to the case. Its commentary specifically identifies unjustified extrapolation and failure adequately to account for obvious alternative explanations as relevant concerns. The proponent must establish the admissibility requirements more likely than not; that is a gatekeeping inquiry, not a guarantee that the opinion is correct. (Legal Information Institute)
An expert who emphasizes platform exposure while ignoring documented prior attempts, acute crises, significant psychiatric symptoms, and substance misuse should be required to explain those omissions.
An expert who moves from a study of depressive symptoms to an opinion about suicide death should explain the inferential steps.
An expert who treats an isolated exchange as representative of months of communications should establish why.
The Supreme Court’s decision in General Electric Co. v. Joiner recognizes that a court may reject an opinion when the gap between the underlying data and the conclusion is too great. An impressive résumé does not supply a missing causal connection. (Legal Information Institute)
This is not an automatic exclusion formula. Reasonable experts may disagree, and not every unresolved alternative defeats an opinion. The defense must meet the same evidentiary standards it invokes against the claimant. (Legal Information Institute)
There is an equally important limit on the defense: multifactorial causation does not mean no one can be legally responsible. California’s concurrent-causation instruction, for example, recognizes that a defendant’s conduct need not be the only cause of harm. Identifying preexisting vulnerability therefore does not automatically defeat a claim of a proven aggravating contribution. The applicable jurisdiction’s law controls. (Justia)
The correct demand is neither corporate immunity nor automatic liability.
It is a supported explanation of the defendant’s alleged contribution, tested against the complete record.
A lawsuit should not be called fraudulent merely because it is weak or unsuccessful. But it should not be called scientifically established merely because its allegations are emotionally powerful.
Compassion for a family and rigor about causation are not opposing commitments. A just legal system needs both.
A verdict is not a treatment plan
The practical danger of scapegoating is that the search for a defendant can displace the search for an explanation.
Dobin and colleagues interpret their findings as pointing toward a more proximal, impulse-driven youth-suicide profile. They emphasize emotional-regulation and problem-solving approaches, together with reducing access to highly lethal means. Their findings do not reduce prevention to identifying which company owned the last screen. (PubMed)
My concern is what gets neglected when that narrower story takes over: the prior warning signs, the developmental vulnerabilities, the acute relationship crisis, the psychiatric deterioration, and the opportunities to interrupt escalation.
Technology belongs in that investigation. It does not deserve to replace it.
Holding a company accountable for demonstrated wrongdoing is compatible with rejecting unsupported blame. Investigating family and psychiatric history is compatible with respecting the child and the bereaved. Recognizing uncertainty is compatible with insisting that preventable harm be prevented.
What is not compatible with serious medicine—or serious law—is deciding on the culprit before reconstructing the case.
Investigate the young person’s life. Examine the actual exposure. Test the alleged contribution.
Accountability requires evidence. Scapegoating requires only a target.
No comments:
Post a Comment