Why Banning Algorithmic Feeds Is the Wrong Fix

Why Banning Algorithmic Feeds Is the Wrong Fix

An off switch sounds tidy. Clean. Heroic.

It is also a distraction. In related news, we also covered: Inside the Starlink Frontline Glitch Nobody Wants to Admit.

When commentators read headlines about lawmakers pushing to shut down algorithmic feeds or hand users a permanent chronological toggle, they usually reach for a familiar moral script. Bad code traps good people. Outrage sells. A big red button sets everyone free.

That story makes great copy. It gets clicks. And it misses almost everything about how modern recommendation engines actually work. The Verge has analyzed this critical subject in great detail.

An "algorithm" is not a single demon sitting in a server room deciding to ruin someone's afternoon. A recommendation system is a multi-stage ranking pipeline—typically a candidate generator paired with a scoring model and a re-ranker—trained to predict which piece of content a given account is likely to click, watch, comment on, share, or return for. When people say "toxic algorithms," they are usually pointing at an objective function and a feedback loop: engagement optimization applied to human attention.

Flipping an off switch on ranking does not flip an off switch on human nature. Nor does it erase the business model paying for the bandwidth.

The Lazy Consensus Around the Off Switch

The standard argument against algorithmic feeds goes something like this: chronological feeds are natural and safe; ranked feeds are manipulative and addictive; therefore, forcing a switch to chronological or disabling recommendation engines fixes the harm.

I have sat in product rooms where teams tried to build "calm" feeds and watched engagement fall off a cliff while the exact same inflammatory accounts kept winning through group chats, repost chains, and external links. I have seen companies blow millions on trust-and-safety theater—cosmetic toggles, warning labels nobody reads, and surface-level feed tweaks—without touching the incentives underneath.

Chronological order is not neutral. It is just a different sorter.

Sort a stream strictly by timestamp and you reward whoever posts most often, posts at peak hours, or automates distribution. In fast-moving news or heated political moments, a raw chronological feed becomes a firehose of repetition and breaking rumors with zero relevance filtering. Worse, if a platform strips out ranking without fixing discovery, new voices struggle to get seen unless they already have a massive follower graph. Legacy accounts lock in their advantage.

Even when users are given a chronological switch, behavior rarely matches the press release. Studies and platform disclosures over the years—including research examining user choices on major social networks—have repeatedly shown that many people switch back to ranked feeds because chronological streams feel messy, repetitive, or disconnected from what they came to see. Convenience wins almost every time.

If a toggle is buried in a menu, almost nobody touches it. If a toggle is front and center and makes the product feel worse, people leave—or turn the toggle back on by Friday.

What an Off Switch Actually Breaks

Before cheering for a button that turns off algorithmic ranking, it helps to look at what happens across the rest of the system.

Discovery and Cold-Start Problems

Imagine a scenario where a video or social platform disables personalized ranking overnight and shows only posts from accounts a user manually follows, ordered newest first. On day one, loyal followers see updates from friends and creators they already know. By day three, anyone who follows fewer than fifty active accounts runs out of fresh material in minutes. Small creators with zero prior audience have almost no path to reach strangers.

Without ranking, discovery shifts from algorithmic recommendation to social graph inertia and search. That helps established media brands and celebrities. It hurts emerging creators, niche communities, and smaller publishers trying to break through without paying for ads.

Moderation at Scale

There is another side to ranking that critics rarely mention: ranking also powers demotion and containment.

Large platforms handle billions of items a day. Human reviewers cannot inspect every post before it goes live. Automated classifiers and ranking signals are routinely used to downrank suspected spam, coordinate visibility limits during fast-breaking events, or reduce the reach of borderline material while classifiers catch up.

Turn off the ranking layer blindly, and you do not just turn off "addictive" recommendations. You can also strip away throttling mechanisms that keep low-quality or harmful content from flooding a feed at full volume.

Ad-Supported Incentives Remain Intact

A ranking model is a mirror held up to whatever metric leadership rewards. If the reward is raw watch time or outrage clicks, the model learns to serve high-arousal content that keeps thumbs moving.

Simply removing personalization does not change the fact that an ad-supported platform makes money when eyeballs stay on screen. If ranked feeds go away, product teams will look for other levers—notifications, streaks, autoplay, creator payouts, or aggressive push alerts—to keep sessions long. Whack-a-mole is not a cure.

Dismantling Common Questions About Algorithmic Feeds

People keep asking the same set of questions whenever a government or regulator targets recommendation systems. Most of those questions start from a shaky premise.

"Do algorithms force people to become extreme?"

Not like a hypnosis beam. Recommendation systems amplify patterns that are already present in human clicks, dwell time, and shares.

Researchers studying polarization and online radicalization—including work from scholars such as Jonathan Haidt on attention economies and youth mental health, alongside empirical media researchers studying echo chambers and algorithmic amplification—point to a messy reality rather than a single villain. Human beings already gravitate toward ingroup loyalty, moral outrage, novelty, and threat. A model tuned purely for engagement notices that angry or shocking posts hold attention longer than calm, nuanced ones, so it serves more of them.

Blaming the math alone lets both platforms and users off the hook. The math is optimizing for a target someone chose. And the user is still the one clicking, pausing, and sharing.

"Would chronological feeds make social media healthier?"

Healthier in one narrow way, maybe—less "infinite rabbit hole" pull on a quiet Tuesday night. Healthier overall? Not automatically.

A chronological feed can still be packed with anxiety-inducing news, status comparison, and conflict if the people and pages you follow post that way. In fact, removing relevance filtering can increase repetitive scrolling as users refresh constantly to see what just dropped a minute ago. Less personalization is not the same thing as less compulsion.

"Can regulators force platforms to make feeds safe with an off switch?"

Mandating a switch or an opt-out is straightforward on paper. Making it work in practice is a regulatory minefield.

Define "toxic" across millions of edge cases, languages, memes, jokes, and political arguments without handing censors a blunt instrument or creating massive false positives. Regulators who focus solely on a switch often end up fighting yesterday's interface while the real levers—ad targeting, retention hooks, notification cadence, and data collection—keep humming underneath.

Where the Contrarian Bet Falls Short

Any honest take on this problem has to admit its own blind spots. Pushing back against "just ban the algorithm" does not mean defending Big Tech's status quo.

First, transparency is still terrible. Independent researchers and the public rarely get clear, auditable visibility into how ranking weights shift or how safety classifiers perform across regions and languages. Saying "it's complicated" is not an excuse for secrecy.

Second, for younger users and vulnerable groups, default settings carry huge weight. Telling a teenager to "just manage your own attention" ignores how persuasive design interacts with developing impulse control. Defaults matter when friction is low.

Third, a blunt off switch is politically easy to understand. Nuanced reforms—auditing objective functions, limiting manipulative dark patterns, separating ad incentives from feed ranking—are harder to explain in a thirty-second clip. Sometimes lawmakers pick a simple lever because complex rules bog down in legal fights.

What Actually Works Instead of an Off Switch

If an off switch is mostly theater, what moves the needle?

  • Change the optimization target, not just the sort order. Stop rewarding pure click-through and rage-bait dwell time. Tie ranking signals to explicit user satisfaction surveys, follow-through behavior, and diverse source exposure rather than raw session length.
  • Give users real control over ranking signals, not just an on/off binary. Let people tune down specific topics, mute keywords or formats, inspect why something appeared, or switch between interest-based, network-based, and chronological modes with zero dark patterns.
  • Audit the business model behind the feed. As long as attention equals ad revenue and retention equals survival, every "safety" feature will face quiet internal pressure to look good without hurting daily active minutes. Real accountability means inspecting incentives, not just interfaces.
  • Invest in media literacy and friction. Deliberate friction—like pausing before resharing unverified claims or separating private chats from public broadcast feeds—does more to cool down pile-ons than rearranging timestamps.

An off switch is a catchy headline. Fixing incentives takes spine.

Stop fighting the sorter and start fixing what you reward.

DG

Dominic Garcia

As a veteran correspondent, Dominic Garcia has reported from across the globe, bringing firsthand perspectives to international stories and local issues.