Gaps
If a slogan needs this list to be empty, the slogan is early
These are not insinuations. They are the measurements this field still owes.
01
Does viewing diagnostic content increase subsequent self-diagnosis?
Why it matters. This is the central untested arrow in public debate. Without it, awareness, selection, and amplification remain observationally equivalent.
What would help. Prospective studies with logged exposure, baseline symptoms, and later self-labelling plus, where ethical, clinical assessment. Quasi-experiments around viral content shocks.
Unknown. Adjacent literatures (cyberchondria, awareness) are not substitutes.
02
Does algorithmic repetition matter more than total screen time?
Why it matters. Most epidemiology still measures minutes. The hypothesized mechanism is concentrated, repeated diagnostic content.
What would help. Exposure traces from recommenders: topic sequences, session depth, and diversity of mental-health content.
Mechanistically plausible; not tested for psychiatric self-concept.
03
Does self-diagnosis increase or decrease professional help-seeking?
Why it matters. The same behaviour could be a bridge to care or a substitute for it. Clinical and platform implications diverge.
What would help. Longitudinal help-seeking studies that distinguish “I think I have X” from “I booked an assessment”.
Mixed suggestions from help-seeking reviews; no decisive social-media-era answer.
04
How often does self-diagnosis agree with clinical assessment?
Why it matters. Accuracy is assumed in opposite directions by competing public narratives. It is an empirical quantity.
What would help. Blinded structured interviews in people recruited from online self-identification, reported by condition.
Unknown as a general fact; likely condition-dependent.
05
Does accuracy differ by condition, age, severity, access, source, and diagnostic complexity?
Why it matters. Autism self-identification in adults is not the same research object as DID self-diagnosis in adolescents or ADHD content on TikTok.
What would help. A programme of condition-specific concordance studies, not a single headline percentage.
Almost entirely open.
06
Can online communities amplify symptom expression without changing underlying disorder?
Why it matters. Public debate collapses signalling, interpretation, and severity. Research rarely unbundles them.
What would help. Designs that measure posting, phenomenological description, and independently rated severity over time.
Plausible; poorly measured. Self-harm literature is the closest analogue.
07
Are effects temporary or persistent?
Why it matters. A 2021 clinic wave, a lasting identity, and a lifelong disorder are different public-health objects.
What would help. Follow-up of FTLB cohorts and of online self-identification cohorts beyond 12–24 months.
Early clinical follow-up exists in specialist centres; population persistence unknown.
08
Which adolescents are most susceptible?
Why it matters. Average null or small effects can hide concentrated risk. Policy aimed at everyone may miss the group that matters.
What would help. Person-specific and risk-enriched designs: anxiety, autism, trauma, loneliness, prior functional symptoms, identity distress.
Heterogeneity is established for well-being; not mapped for self-diagnosis pathways.
09
What role does belonging play?
Why it matters. Belonging can be a social cure or an illness-identity trap depending on group norms.
What would help. Social-identity measures inside diagnosis-specific communities, linked to coping and impairment.
Strong theory; thin platform-era data.
10
What role does loneliness play?
Why it matters. Loneliness could drive both platform use and diagnosis-seeking. It is a confounder and a possible mechanism.
What would help. Models that treat loneliness as exposure, mediator, and outcome—not a nuisance covariate only.
Widely invoked; rarely isolated.
11
How important are recommendation algorithms, as opposed to search, friends, and culture?
Why it matters. If the driver is culture, ranking tweaks will not do what reformers hope. If ranking is the driver, design is a public-health lever.
What would help. Comparisons of recommended vs requested content; platform natural experiments.
Public understanding of rankers is better than mental-health outcome evidence.
12
Can social media produce functional symptoms in susceptible people?
Why it matters. This is the strongest version of the harm claim and the easiest to overstate from FTLB clinics.
What would help. Careful FND phenotyping, exposure measurement, and competing-risk models including anxiety and pandemic stress.
Clinical association for FTLB; causation unproven; generalization forbidden without new data.
13
Can it exacerbate existing symptoms?
Why it matters. A weaker, more probable claim than “creates disorders,” with different clinical advice (content hygiene vs disbelief).
What would help. Within-person symptom trajectories around exposure bursts in diagnosed cohorts.
Suggestive in eating-disorder and self-harm literatures; incomplete elsewhere.
14
Can social media improve outcomes through earlier recognition?
Why it matters. If yes, crude crackdowns on mental-health content could harm the people services already miss.
What would help. Outcome studies: time to assessment, treatment uptake, and functioning after online recognition.
Plausible and partially supported; not quantified for short-form video.
15
What distinguishes beneficial communities from harmful ones?
Why it matters. The policy-relevant unit may be norms and moderation, not “online mental health” as such.
What would help. Comparative ethnography plus outcome measurement: recovery talk vs symptom competition; moderation; off-ramps to care.
Clinicians have impressions; systematic distinctions are still thin.