Critiques of social media guide: sorting real complaints

Someone searching for “critiques of social media guide” needs a practical distinction: complaints can target a design choice, a business-model consequence, or a property of communicating at scale. Some of what people object to is a choice somebody made and could unmake. Some of it follows from communicating through a medium that copies perfectly, and no design decision removes it.

Sorting a complaint into the right category does not tell you whether it is right. It tells you what a solution would have to look like, and it stops people demanding fixes that cannot exist while ignoring the ones that could.

What to take away

The three categories

The middle row is where the real argument lives, and it is genuinely an argument, since the alternatives all have costs and the costs fall on different people.

Sorting a complaint

Take the complaint, and ask three questions in order.

Real critiques sorted

Take the filter-bubble critique associated with Eli Pariser: when the complaint is that a platform's personalization narrows what a user sees, sort it as a design choice. A service can change its personalization without ceasing to be a social network; the complaint is not that communication itself inherently filters information.

Surveillance capitalism is a business-model critique. Shoshana Zuboff's 2018 book describes Google and Meta among the digital companies it examines, and argues that their business models represent a new form of capitalist accumulation called surveillance capitalism.

The attention-economy critique is a business-model complaint. Tim Wu's The Attention Merchants traces the trade in human attention, and the argument is that a service paid for by advertising has to keep people in front of it. Change who pays and the pressure changes.

Sherry Turkle's Alone Together sorts as a property of the medium. Her complaint is that we expect more from technology and less from each other, and that connection starts to stand in for conversation. That would hold in a service with no advertising, no ranking and no company behind it.

Jaron Lanier's Ten Arguments for Deleting Your Social Media Accounts Right Now is mostly a business-model critique: the money comes from modifying behavior, so the manipulation is the product rather than a malfunction. The parts that name a feed or a notification split off as design choices.

Tristan Harris and the Center for Humane Technology, and the documentary The Social Dilemma, make both complaints at once. Attention extraction is the business-model half. Infinite scroll, autoplay and the notification badge are design choices a competitor could drop next week.

danah boyd's work on networked publics and context collapse is a property-of-the-medium complaint. A message reaches an audience it was not written for because copying is free and the audience is invisible.

Cathy O'Neil's Weapons of Math Destruction targets opaque automated scoring, a design choice a competitor could make visible. Siva Vaidhyanathan's Antisocial Media targets the dependence of public conversation on one advertising-funded company, a business-model consequence. Cal Newport's Digital Minimalism treats attention capture as a personal discipline, which is what a complaint becomes when nobody can point to a switch that would end it.

Frances Haugen's disclosures, published as the Facebook Files, are evidence rather than an argument, and they land in the business-model category: internal research about what engagement-based ranking did to the people using the service.

The complaint that copied misinformation can travel faster than corrections even without ranking or advertising is a property-of-the-medium claim. If the complaint instead targets a platform's recommender amplifying it, that is also a design-choice complaint; if financial incentives reward the amplification, the business model is part of the diagnosis too.

The third test is the one people skip. Surprisingly many complaints survive it: material spreading faster than corrections, context not traveling with a message, small groups overwhelmed when found, and anything written down staying permanent.

Those happen wherever many people can publish to many people. They happened in every earlier form of this too, which the record of early social networks shows.

Why the categories get confused

Because the same visible symptom can come from any of the three, and the symptom is what people see.

Something spreads widely and does harm. That could be a ranking decision, which is a design choice. It could be that the ranking is built to maximize engagement, which is a business-model consequence.

Or it could be that copying is free and correction is not, which is a property of the medium. Usually all three occur at once, in some proportion nobody can measure from outside.

The practical consequence is that a complaint which does not say which category it is in cannot be answered. It gets a response addressing one category, the response is correct and insufficient, and everyone concludes the other side is arguing in bad faith. Sorting critiques of social media into families that share a category is what makes those arguments move.

What each category implies about remedies

Design choices are answered by design changes, and they are answered often. The lesson from watching this is that a design change is real and small: it fixes the specific thing and the underlying pressure reappears in another feature. That is not cynicism; it is what happens when the pressure is in a different category from the fix.

Business-model consequences are answered by changing what pays for the service, which means subscriptions, public funding, a different scale of operation, or regulation. The instruments have names. In the United States, Section 230 of the Communications Decency Act sets what platforms are liable for, the Children's Online Privacy Protection Act governs what services may collect from children, and the Federal Trade Commission enforces against deceptive practices; in the European Union, the Digital Services Act imposes duties on large platforms. Every one of those has a distributional cost, and the argument is about who bears it, which is a political question and not a technical one. The general framing of costs falling on people outside the transaction is externality.

Properties of the medium are answered by local mitigation, and that is all. Smaller spaces, slower norms, gates on membership, moderation. None of it scales, which is why the useful work happens inside internet communities rather than in policy.

Where the sorting is genuinely hard

Two cases resist the test.

Amplification. Whether ranking counts as design or as business model depends on what the ranking is for, and outsiders cannot see that. The honest position is that it sits between the categories and that a claim placing it firmly in one is making an assumption.

Scale itself. Whether the size of a service is a choice or a property is contested. A service can decide to grow, so it looks like a choice. Growth is also the thing the network effect produces without anyone choosing it, so it looks like a property. Both readings are defensible and the disagreement is real.

Teen mental health. Jonathan Haidt's The Anxious Generation makes the sharpest version of this complaint, and it is the clearest case of a contested category. Read as the effect of engagement-optimized feeds on minors, it sorts as design and business model. Read as the effect of mediated social life on how children develop, it is a claim about the medium.

Saying which cases are hard is part of a decent argument. A sorting scheme that classifies everything cleanly is not describing this subject.

Common questions

Does calling something a property of the medium excuse it?

No. It says the remedy is not a redesign, which usually means the remedy is smaller spaces and better norms. That is a harder answer, not a lighter one.

Are the categories stable?

Mostly. A design choice can become a business-model consequence once revenue depends on it, and that transition is worth watching, since it explains why some easy fixes stop happening.

Can a complaint be in two categories?

Frequently, and the useful move is to split it into the part with a design fix and the part without. Arguments improve immediately when someone does this.

Which category should I put my own complaint in?

Run the third test first.

If it would happen anyway with no company involved, you learn the most important thing about it before other work. Testing a claim against its plainest rival is the same habit used to date a feature on platform timelines.