Designing for Substance: Platform Incentives and the Attention Economy

Before food-safety laws, a factory could sell almost anything that fit in a can. Before tobacco warnings, a company could market cigarettes to children. Before seatbelts were standard, carmakers could treat safety as the driver’s private responsibility. In each case, the design of a product was treated as a purely commercial decision until the public costs grew too large to ignore. Then the frame shifted: some design choices are not merely business choices; they are civic choices.

We are at a similar shift with attention-extracting platforms. The question is no longer whether they are legal or popular. It is whether their default designs are compatible with a healthy society. And if they are not, what would it take to build something better?

Point C1 The attention economy is not an inevitable feature of the internet. It is the product of specific design choices—engagement metrics, personalized recommendations, infinite scroll, and advertising-based business models—that could be redesigned.

The Attention Economy Is a Design Choice

It is easy to talk about “the attention economy” as if it were a natural force, like gravity or weather. People spend hours on platforms because platforms are addictive, and platforms are addictive because that is what works. This story makes the current shape of the internet feel inevitable.

It is not. Engagement optimization is a choice driven by business models, not a law of nature. A platform that sells advertising against time spent has every reason to maximize time spent. A platform that sells subscriptions, or that is funded by a community, or that is run by a public institution, has different incentives. The metrics a company optimizes for are not discovered in a physics lab; they are chosen by people in rooms.

Research supports this. A PNAS Nexus study by Milli and colleagues found that engagement-based ranking can amplify divisive content even when users say they prefer less divisive material. The system is not simply giving people what they want; it is giving people what keeps them watching, clicking, and scrolling. Yale research showed that likes and shares teach people to express more outrage over time. The feedback loop is not accidental. It is the product of a design that rewards emotional escalation because emotional escalation produces engagement.

This matters because design choices can be changed. A car company can choose to install seatbelts before it is required to. A food producer can choose cleaner processes before a scandal forces its hand. A platform can choose different defaults, different metrics, and different business models. The question is not whether alternatives are technically possible. The question is whether the incentives can be shifted so that the better choice becomes the easier choice.

Learning From Earlier Design Failures

History offers several useful analogies, not because the internet is exactly like tobacco or food production, but because each case shows how a product’s design went from “private business” to “public problem” to “shared civic responsibility.”

Tobacco regulation is the most often cited. For decades, cigarettes were marketed as freedom, sophistication, and even health. The product design—nicotine delivery, marketing imagery, flavors—was treated as a commercial decision. Then the public-health costs became visible. Warning labels, marketing restrictions, and youth-access rules did not end smoking, but they shifted the default. A product once celebrated became a product that had to answer to public health. The design itself became a civic question.

Food safety offers a different lesson. Before public standards, producers had strong incentives to cut corners: dilute milk, use unsafe preservatives, hide contamination. Markets alone did not fix this, because buyers could not easily inspect every can and bottle. Public standards changed what “normal” production looked like. The result was not a ban on food; it was a floor beneath which no producer was allowed to fall.

The shift from clickbait to subscription journalism shows a market-level change. When advertisers paid by the click, the metric was the click. Headlines grew louder, facts grew softer, and trust became a side effect rather than a goal. When readers began paying directly—through subscriptions, memberships, and patronage—the metric shifted toward trust over time. The business model did not guarantee quality, but it made quality a viable strategy rather than a charitable one.

Open-source software offers yet another model. Much of the internet’s infrastructure runs on tools developed for reputation, utility, and community rather than engagement optimization. The incentives are different: recognition among peers, solving a real problem, building something durable. The result is inspectable, modifiable, and often longer-lived than commercial alternatives. It is proof that a different incentive structure can produce different design outcomes.

None of these analogies maps cleanly onto social media. But together they suggest a pattern: when a product’s private design creates large public costs, societies eventually develop new accountability structures. The attention economy is now reaching that point.

What Alternative Metrics Could Look Like

Point C2 Alternative metrics such as time well spent, creator sustainability, learning outcomes, civic value, and long-term user flourishing can be measured and rewarded, though none is perfect or easy to game-proof.

If engagement is not the only possible metric, what else could platforms optimize for? Several candidates have been proposed, and each comes with tradeoffs.

Time well spent. Instead of maximizing minutes on site, a platform could aim for sessions that users retrospectively judge as valuable. This was famously proposed as a direction for Facebook’s design team. The challenge is measurement: asking users afterward is expensive and biased, and passive signals like dwell time do not reliably distinguish deep reading from distraction.

Creator sustainability. A platform could reward creators whose work supports their livelihood without requiring constant viral performance. This shifts attention from the single hit toward a longer arc. The risk is that “sustainability” can be defined in ways that favor incumbent creators or that obscure how income is distributed.

Learning outcomes. Educational platforms have long optimized for what users actually learn. The approach is harder to apply to general social media, where the goal is not a curriculum but a healthier information environment. Still, the principle matters: a platform could ask whether users leave knowing more true things than when they arrived.

Civic value. A platform could elevate content that helps people participate in public life—informed discussion, local information, civic organizing. The danger here is paternalism: whoever defines “civic value” has power, and that power needs scrutiny.

Long-term flourishing. This is the broadest and hardest to measure. It asks whether the platform contributes to users’ health, relationships, and sense of agency over months and years. It is easy to dismiss as fuzzy, but it is also the standard most people would use if asked whether a product is good for them.

No metric is magic. Every one can be gamed, and every one requires judgment about what counts. But the existence of alternatives undermines the claim that engagement is the only thing platforms can optimize. The hard problem is not imagining alternatives; it is creating the incentives and accountability that make them stick. The Alternative Metrics article explores these candidates in more detail.

Accountability From Three Directions

Point C3 Accountability for platform design can emerge through regulation, public pressure, and user migration to better-designed platforms. No single lever is sufficient on its own.

Changing incentives requires pressure from more than one direction. Three stand out.

Regulation. The European Union’s Digital Services Act is a recent example. It requires large platforms to assess and mitigate systemic risks, including risks to mental health and democratic discourse. It demands transparency about recommendation algorithms and advertising. It is not a blueprint for a substance-friendly internet, but it creates legal pressure to treat design as a public-interest question rather than a purely private decision.

India’s Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021, amended in 2026 with a deepfake-labelling requirement and other provisions, take a different approach. They impose due-diligence requirements on intermediaries, including grievance mechanisms, content takedown procedures, and obligations around harmful or misleading material. Like all such rules, their effectiveness depends on implementation, enforcement, and the balance they strike between accountability and overreach. But they show that governments are no longer treating platform design as hands-off territory.

Regulation has limits. It can be slow, captured by industry, or weaponized against speech. Done poorly, it can entrench large incumbents who can afford compliance while excluding smaller competitors. But done well, it can set floors—like food-safety standards—that make the worst designs costlier and the better designs more competitive.

Public pressure. Regulation is rarely the first mover. Public pressure—from researchers, journalists, civil society, and users—often creates the conditions for it. The Center for Humane Technology and similar groups have helped reframe attention extraction as a design problem rather than a personal-failure problem. Investigative reporting has documented harms. Academic research has supplied evidence. Whistleblowers have made internal choices visible. None of these alone changes a platform, but together they shift what is socially acceptable.

Public pressure also works through internal channels. Employees at major platforms have pushed for ethical review of products. Shareholder proposals have asked companies to measure harms. Consumer boycotts and user campaigns have forced policy changes. The common thread is that design choices become public controversies, and public controversies create costs. See Public Pressure and Internal Accountability for a fuller account of these channels.

User migration. The third lever is the oldest one in the market: people leaving for something better. This is where product alternatives matter. If users have somewhere else to go that respects their attention, the threat of migration becomes real. Without an exit, regulation and protest have less bite. User Migration and the Exit Problem examines the switching costs and network effects that make exit hard.

Migration is hard. Network effects lock people in. Friends, employers, and public institutions are often on the dominant platform. Leaving can mean social and professional cost. But history shows that network effects are not permanent. Email, the web itself, and mobile messaging all saw shifts once the alternative was good enough. The fediverse and other decentralized models are attempts to build that alternative for social media.

Substance-Friendly Design Is a Category, Not One App

Point C4 Substance-friendly design is a category that includes chronological feeds, user-chosen algorithms, expert curation, reputation systems, and friction against virality. It is not limited to a single product or business model.

If the goal is to help users leave better than they arrived, several design patterns move in that direction. Friction, Chronological Feeds, and User-Chosen Algorithms explores these patterns in more detail.

Chronological feeds. A simple feed ordered by time removes one of the most powerful optimization levers: the algorithm that selects whatever maximizes engagement. Users see what their contacts posted, when they posted it. The feed may be less addictive, but it is also more predictable and less manipulable.

User-chosen algorithms. Instead of a single opaque recommendation system, a platform could let users choose or configure what they see. One user might prioritize depth; another might prioritize local news; another might want a purely chronological view. Transparency and choice do not guarantee good outcomes, but they distribute control more fairly than a hidden default.

Expert curation. Some platforms blend algorithmic sorting with human editorial judgment. This can mean librarians, journalists, moderators, or community stewards who have a name, a reputation, and accountability. Expert curation is expensive and can be biased, but it introduces a form of judgment that pure engagement optimization lacks.

Reputation systems. Instead of counting likes, shares, and followers, a platform could signal reliability, helpfulness, or contribution over time. Reputation systems have their own failure modes—gaming, cliques, status anxiety—but they can reward behavior that benefits the community rather than behavior that captures attention.

Friction against virality. The fastest-spreading content is not always the most valuable. Some platforms have experimented with cooling-off periods, sharing prompts, and reduced amplification for highly emotional posts. The goal is not to stop sharing but to add a small pause between reaction and distribution. A moment of friction can reduce the spread of misinformation and outrage without requiring a central censor.

The fediverse, including Mastodon, is a working experiment in several of these ideas. Research on Mastodon administrators and feed curation has documented both the promise and the difficulty of decentralized social media. The promise is user control, community governance, and escape from a single corporate algorithm. The difficulty is moderation at scale, funding sustainability, technical barriers to entry, and the fragmentation of communities across instances. The fediverse is not a finished answer, but it is a real category of alternative.

The Better Question

Point C5 A substance-oriented platform would ask how to help users leave better than they arrived, rather than how to maximize time on platform.

This question sounds simple, but it inverts the usual product logic. Most platforms measure success by retention: daily active users, session length, return visits. A substance-oriented platform would measure success partly by what happens after the user leaves. Did they learn something? Did they connect with someone? Did they act on information that mattered to them? Did they feel more capable rather than more depleted?

This is not a call for platforms to become wellness apps. Entertainment, distraction, and social play are legitimate human needs. But there is a difference between a product that sometimes distracts and a product whose entire business model depends on maximizing distraction. The first is like a bookstore that also sells coffee and magazines. The second is like a casino that designs every surface to keep you at the table.

The question also changes how product teams think about growth. Growth by extraction and growth by trust are different strategies. Extraction growth asks: how do we get one more minute? Trust growth asks: how do we become the kind of place users recommend because it made their lives better? Both can produce revenue. Only one is compatible with a healthy public sphere.

Life-Phase Threads

What would this look like across different life stages?

Children. Age-Appropriate Design means no infinite scroll, no autoplay, no dark patterns, and no algorithmic recommendations before a certain age. A child’s feed should be bounded, predictable, and controlled by caregivers or educators. This is not a radical idea; it is the digital equivalent of child-proofing.

Students. Educational and civic content should get algorithmic priority; entertainment should be available but not the default. A learning environment does not need to ban fun, but it should make the useful path at least as easy as the distracting one.

Workers. Focus-protective defaults—batch notifications, no infinite feeds during work hours, clear session boundaries—would treat attention as a resource the user is trying to protect, not a resource the platform is trying to mine.

Elderly users. Safety and verification should be prioritized. Sharing should be slowed by gentle friction that makes misinformation harder to spread. Scams and manipulative content target older users in particular; design that protects them is design that protects everyone.

What This Article Is Not

It is worth repeating the boundaries. This article is not a business plan for a new platform. It is not a regulatory handbook. It is not a utopian blueprint. It is an exploration of plausible directions.

Nor is it a claim that regulation or alternative platforms will solve everything. The attention economy is deeply entangled with advertising markets, political polarization, economic inequality, and human psychology. No design change will untangle all of that. But design is one of the levers, and it is a lever we can pull.

Open Questions

• Which alternative metrics are robust enough to survive gaming and commercial pressure?
• How do we design regulation that protects users without empowering censorship?
• Can decentralized platforms like the fediverse reach the scale and usability needed for mainstream migration?
• What business models make substance-friendly design financially sustainable?
• How should age-appropriate design be enforced across jurisdictions?
• When does friction against virality become friction against legitimate speech?

These questions do not have easy answers. But asking them is the first step toward making substance the easier path.

Related in This Series

• The Attention Extraction — the diagnosis and thesis of the series.
• The Generational Bet — why the AI moment raises the stakes for design choices.
• Engagement Is a Design Choice — why ranking metrics are not inevitable.
• Regulation as a Floor — what India’s rules and global frameworks can and cannot do.
• A Map of Levers — who can pull which levers to shift the default.
• Business Models That Reward Substance — subscriptions, patronage, public funding, and other alternatives to attention-based advertising.
• Attention, Substance, and the AI Moment — the full series guide and reading paths.