Feeds are triaging patients by engagement, and the proposed defence is literacy, not moderation

Live Science has published an interview with Deborah Cohen, a broadcaster and journalist with a medical degree from the University of Manchester, about her book "Bad Influence: How the Internet Hijacked Our Health," published by Oneworld in 2026 ^[1][12]. The most consequential line is the publication's own framing rather than any of Cohen's answers: algorithms deliver users an individually tailored stream of content, promoting and prioritising the most engaging posts, with little oversight ^[2].

Read that as an operating description of a triage layer. Something decides which account of a symptom a person encounters first, and the sorting criterion is engagement, with no oversight function named ^[2]. The same channel carries genuine support groups and symptom information alongside actors promoting products, spreading misinformation and preying on fears, so the ranking is not filtering between those categories in any way the source describes ^[3]. The interview is headlined with Cohen's observation that "There's a sense that these algorithms are objective and they get to know you" ^[4], which is the failure mode that matters: a ranking layer read as a diagnostic one.

Cohen's account of the downstream cost is about expectations rather than facts. She argues that ideas, concerns and expectations shape health, and that a consultation is largely the clinician establishing what those are ^[5]; those expectations are set by the information environment ^[6]. Her stated concern is the mismatch that follows. Social media supplies certainty and quick fixes, while the NHS cannot supply certainty and instead offers complex solutions, or none, with competing risks, benefits and harms to balance ^[7]. On that account the harm does not land in the feed. It lands in the appointment, as disappointment.

The availability asymmetry compounds it. Cohen notes that the technology is there 24/7 and that large language models can be queried at any hour, and asks what that does to trust ^[8]. She expects LLMs to change the information environment again and produce what she calls a new wave of health literacy ^[6]. Asked whether social media is widening the doctor-patient gap or filling it, she says both, and that "the gap is filled, but not necessarily by the right things," asking whether support has to be on TikTok and whether there are other, safer ways to do it online ^[9]. She also concedes the demand side: an abundant health service, which she says no country has, would need these tools less ^[10].

Note where the fix is aimed. The interview identifies unoversighted engagement ranking as the mechanism ^[2] but positions health literacy as the response ^[6], which puts the burden on the person being triaged rather than on the system doing the triage ^[17]. That is a defensible read of what is actually enforceable, and it is also a lot to ask of a patient at 2am.

This is one clinician-journalist's argument in a book interview, shortlisted for the 2026 Royal Society Trivedi Science Book Prize ^[13], and it carries no prevalence or effect-size figures ^[18]. Watch whether chat interfaces displace the feed as the first-contact layer ^[8], whether any health service builds the safer forums Cohen gestures at ^[9], and whether self-diagnosis patterns such as the ADHD case her excerpt covers ^[15] ever get measured rather than described. Cohen calls the whole thing a global experiment in public health that will take years to understand ^[11]; experiments without instrumentation take longer.