The Designer Isn't the One Who Made Your App Addictive. The Dashboard Is.
In February 2026, the European Commission issued a preliminary ruling against TikTok citing infinite scroll, autoplay, push notifications, and hyper-personalized recommendations as deliberate addiction vectors under the Digital Services Act. In July, Meta got the same treatment. Both cases describe the interface as the offense — the scroll mechanic, the notification timing, the feed ranking, named individually as design choices that manipulate rather than serve. Read the coverage and you'd think some designer, somewhere, sat down and decided to build something addictive on purpose, out of malice or indifference to the people using it.
That's not what happened, and treating it like what happened is exactly why the enforcement wave won't change the next product cycle. The infinite scroll wasn't a rogue decision by an unethical designer. It was the correct answer to the question the org actually asked, which was never "how do we serve this user well" — it was "how do we increase session length," because session length is the number that gets reported to the board, tied to ad revenue, and attached to the designer's own performance review. You can't ethics-train your way out of an incentive structure. You can only build the thing the incentive structure rewards, or lose your job to someone who will.
The Metric Was Never Neutral
Engagement metrics get talked about as if they're a value-free instrument — you measure DAU, session length, scroll depth, and then separately decide whether to act ethically with what you learn. That framing doesn't survive contact with how product teams actually operate. The metric isn't a measurement taken after the fact. It's the design brief. If a team is graded on session length, every interface decision gets filtered through "does this increase or decrease time spent," which means autoplay wins the A/B test against a stopping cue every single time, because autoplay is structurally better at the thing being measured, independent of whether it's better for the person using the product. The metric doesn't get gamed by bad actors. It gets satisfied exactly as designed, by good designers doing their jobs competently, which is the uncomfortable part nobody wants to sit with.
This is the mechanism the EU's own enforcement language gestures at without quite naming. The DSA rulings describe patterns — infinite scroll, autoplay, notification cadence — as if they were independently chosen manipulation tactics. They're not independent choices. They're the convergent output of any team optimizing the same metric, which is why nearly every major platform arrived at structurally identical interfaces despite having no coordination with each other. Convergent evolution isn't collusion. It's what happens when every organism is solving for the same fitness function.
This Is a Different Problem Than "Dark Patterns Are Now a Legal Liability"
The legal exposure itself is real and growing — the fines started landing in 2026, and Article 25 plus FTC click-to-cancel enforcement means the cost of shipping manipulative patterns is no longer theoretical. But legal liability changes what companies are willing to risk getting caught doing. It doesn't change what companies are still measuring internally, which means it changes the visible dark patterns — the confirm-shaming button, the maze-like cancellation flow — faster than it changes the structural ones, like a feed ranking algorithm tuned purely for engagement, which is much harder to point at in a regulatory filing because it doesn't look like a trick. It looks like "personalization."
That's the gap regulation can't close by itself. You can outlaw the cancel button that requires four confirmation screens. You cannot easily outlaw "the recommendation algorithm is very good at knowing what keeps you scrolling," because that's indistinguishable, on paper, from the recommendation algorithm just being good at its job. The fines are real and the accountability is overdue. But they're aimed at the symptom that's easiest to litigate, not the metric sitting upstream of it, still fully legal, still on every quarterly dashboard.
What Changes When the Metric Changes
The counter-evidence for this argument is instructive, not theoretical. Products that ship a "time well spent" style metric — a completion signal, a satisfaction score measured after a session rather than during it, a retention number weighted against a user's own stated goals — produce visibly different interfaces than products optimized purely for session length. Not because the designers on those teams are more ethical. Because the thing they're being graded on rewards a different outcome, and design, like any optimization process, faithfully produces whatever the reward function actually asks for. Swap the metric and the interface changes without a single conversation about ethics ever needing to happen, which is the strongest evidence available that the interface was never really the variable in the first place.
This is also why "responsible design" frameworks that stop at guidelines — checklists, principles, a values statement in the design system documentation — tend to produce so little measurable change. A checklist competes against a KPI that's tied to someone's bonus, and the KPI wins essentially every time, because the checklist is aspirational and the KPI is load-bearing. Anyone who's sat in a roadmap meeting where a "delight" feature loses its budget to a growth experiment with better projected engagement numbers has watched this play out in real time. The checklist doesn't have a number attached. The growth experiment does.
The Actual Fix Skips the Design Team Entirely
If the goal is fewer addictive interfaces, the leverage point isn't a design ethics workshop, and it isn't even, primarily, the regulatory fine — though the fine helps by making the downside real. It's changing what gets measured before a single wireframe exists. A product org that reports retention-weighted-by-user-outcome to its board will ship different interfaces than one reporting raw session length, not because its designers have different values, but because they're solving a different equation. Regulators fining the output are treating a systems problem as a conduct problem, and conduct problems get fixed with training while systems problems get fixed by changing what the system rewards.
The designer who built the infinite scroll wasn't the villain in this story, and pretending otherwise lets the actual decision — what the org chose to measure, and who signed off on that choice above the design team's pay grade — stay exactly where it's always been: off the record, off the regulatory filing, and firmly in place for whatever ships next.
Cover photo by Towfiqu barbhuiya via Pexels.