Holding platforms accountable for algorithms that promote self-harm is hard because research linking mental health and social media doesn't establish causation
Context & Ripple Effects
This closes a loop that has been building for years. The critique started with amplification rather than censorship — back in 2018, Wired argued the real issue was opaque algorithms recklessly amplifying harmful content, not political bias. Then came the policy vacuum: Stanford researchers found during the COVID-19 mental health crisis that most tech platforms had no policies at all on self-harm discussions.
By June 2022 the stakes turned legal, when Meta faced eight lawsuits claiming its algorithms contributed to attempted suicide and eating disorders in young people. Today's piece identifies the weak point in those claims: the research base linking social media to mental illness shows association, not causation — and earlier calls for platforms to build in-house mental health expertise only matter if the causal chain can be demonstrated.
First-order effects
- Plaintiffs in the Meta lawsuits must win on correlational evidence alone, giving defense counsel a ready-made argument that association between algorithmic exposure and self-harm does not prove the algorithm caused it.
Second-order effects
- With external studies unable to settle causation, pressure shifts to platform-internal data — making the in-house mental health knowledge advocates pushed for in 2019 both a potential evidence source and a liability exposure for the companies holding it.
Third-order effects
- If litigation keeps stalling on causal proof, accountability frameworks may pivot from outcome-based claims toward conduct-based ones — judging how recommendation systems amplify harmful content operationally, the standard the 2018 amplification critique implicitly demanded.
The trend: Platform accountability is migrating from what companies host to what their algorithms amplify, and the absence of causal evidence is becoming the decisive battleground in that shift.