Blog · Review Essay · Last reviewed June 25, 2026

The Platform Society and the Public Values Inside the Interface

Jose van Dijck, Thomas Poell, and Martijn de Waal's The Platform Society is one of the cleanest maps of the world that now sits between ordinary life and public institutions. Look past social media and app stores and the real subject appears: the deeper migration of news, transport, health, education, labor, and public knowledge into privately governed digital infrastructures. Read in the AI era, the book explains why model platforms have stopped being mere tools and become the administrative environments through which speech, work, learning, care, and delegation are made actionable.

For this review, a platform society means a social order in which services that look like optional apps become operational infrastructure for public functions. The key question is not whether a platform is convenient. It is whether the values attached to news, mobility, health, education, work, speech, safety, and democratic accountability survive when those functions are routed through private rules, metrics, ranking systems, and data flows.

The sharper test is evidence. A public value has not survived merely because a company names it in a policy or a regulator names it in a statute. It survives when the system leaves usable records, limits extraction, supports appeal, permits independent scrutiny, and can be exited without destroying the public function it has absorbed.

A platformized public function should therefore be reviewed as infrastructure: name the governed population, the decision log, the data boundary, the public-law hook, the appeal route, and the exit plan before the interface becomes the institution's normal way of seeing.

The Book

The Platform Society: Public Values in a Connective World was published by Oxford University Press in 2018. Oxford Academic lists the online edition as published on October 18, 2018, the print edition as available on November 29, 2018, DOI 10.1093/oso/9780190889760.001.0001, online ISBN 9780190889807, and print ISBN 9780190889760. Utrecht University's research portal records the book as a 2018 Oxford University Press academic book of 240 pages under the subtitle Public Values in an Online World.

The book's central claim is that platforms have moved into the heart of social organization. They do not simply host interactions that would otherwise happen elsewhere. They structure those interactions through accounts, feeds, rankings, reputation systems, payment rails, data extraction, moderation, metrics, personalization, and rules that are usually easier to obey than to inspect.

That makes the book useful for thinking about more than platform companies. It is a book about institutional replacement. When a service promises to make some sector more efficient or more connected, it may also be moving authority out of public institutions and into private infrastructure.

The important unit is the platformized function, not the brand. A newsroom, classroom, clinic, taxi network, benefits office, public archive, or research database can be changed by a platform even when the interface looks modest. The test is whether the institution still controls its mission, records, standards of evidence, professional judgment, and exit path after the platform becomes normal.

Current Context

As of June 25, 2026, the book's public-values question has become an operating regulatory question. The European Union's Digital Services Act applies a tiered regime to online intermediaries, hosting services, platforms, marketplaces, very large online platforms, and very large online search engines. The European Commission says very large platforms and search engines are services with more than 45 million monthly users in the EU and face the DSA's strongest obligations around systemic-risk assessment, independent audit, advertising transparency, recommender-system choice, mitigation duties, and researcher data access.

The DSA also makes the evidentiary layer more visible. The Commission's DSA Transparency Database collects platform-submitted statements of reasons for moderation decisions and says it enables near-real-time tracking of those decisions. In July 2025, the Commission adopted a delegated act on data access so vetted researchers can seek internal data from very large platforms and search engines to study systemic risks and mitigations; the European Centre for Algorithmic Transparency says the delegated act entered into force on October 29, 2025 and researchers can submit applications through the Data Access Portal. Those tools are imperfect, but they turn platform power into records outsiders can challenge.

The neighboring Digital Markets Act treats large gatekeepers as market bottlenecks, not merely popular services. The Commission describes the DMA as a law for fairer and more contestable digital markets, with criteria for gatekeepers that provide core platform services such as search engines, app stores, and messaging services; its gatekeepers portal lists 23 designated core platform services as of the current page reviewed here. That distinction matters for The Platform Society: public values can fail through monopoly and dependency as much as through a single harmful post or privacy breach. If an app store, search engine, messaging service, ad system, cloud service, or model platform becomes the required route to users, then contestability, interoperability, portability, and exit are public-value questions.

U.S. regulators have also moved the debate from abstract platform power to evidence about data practices. In September 2024, the FTC staff report on social media and video streaming services described extensive surveillance, weak privacy controls, inadequate protections for children and teens, broad data sharing, retention problems, and limited user control over how personal data fed automated systems. That report is not a general platform statute, but it supports the book's diagnosis: platformization turns everyday participation into institutional data supply.

The United Kingdom's Online Safety Act implementation shows the same movement from platform norms to enforceable duties. Ofcom's illegal-harms materials, published December 16, 2024 and updated through June 25, 2026, require providers to assess illegal-content risks and take the safety measures in the Codes of Practice or use other effective measures. Ofcom's 2026 updates address intimate image abuse, crisis protocols, and new priority-offence changes for serious self-harm and cyberflashing. That matters because safety governance is no longer only a trust-and-safety team deciding case by case; it is becoming a documented system of risk assessment, mitigation, recordkeeping, and enforcement.

AI regulation adds a second layer. The EU AI Act entered into force in 2024. The AI Act Service Desk timeline says general provisions, AI-literacy duties, and prohibitions applied from February 2, 2025; general-purpose AI rules and EU-level governance from August 2, 2025; the majority of rules, including Annex III high-risk rules and Article 50 transparency rules, from August 2, 2026; and high-risk rules for AI embedded in regulated products from August 2, 2027. NIST's voluntary AI Risk Management Framework and Generative AI Profile give a nonbinding U.S. risk-management vocabulary. Together, these sources do not solve the platform society problem, but they make it harder to claim that model platforms are only neutral tools.

The Platform Mechanisms

The book is especially strong because it names platform power at the level of mechanism. The key mechanisms are datafication, commodification, and selection.

Datafication turns activity into records: location traces, clicks, ratings, searches, transactions, biometric signals, social ties, learning behavior, work histories, and health disclosures. Commodification turns some of those records, relations, and dependencies into revenue streams. Selection determines visibility, ranking, recommendation, exclusion, and priority.

Those mechanisms matter because they show how platforms make reality recursive. A user acts inside a system. The system records the action. The record shapes future ranking, pricing, eligibility, or recommendation. The user adapts to the changed environment. The adaptation becomes new evidence. Over time the platform does not merely observe a social world. It helps produce the world it later claims to measure.

The tighter formulation is this: platformization converts activity into data, data into leverage, leverage into ranking, and ranking into behavior. That is why the book belongs beside the site's work on recommender systems, algorithmic transparency, data minimization, opaque scoring systems, and digital identity. The same loop appears in feeds, ride-hailing, learning analytics, health tracking, marketplace trust scores, and AI assistants.

Mechanism is also where accountability becomes concrete. A public-value review should ask which action is being datafied, what retention rule governs the trace, who can reuse it, which ranking or model consumes it, which downstream decision it affects, and what remedy exists if the record is wrong. Without that chain, values remain slogans attached to a black box.

Public Values Under Private Design

The book's most important phrase is "public values." Oxford's abstract identifies values such as privacy, accuracy, safety, security, fairness, accessibility, democratic control, and accountability as stakes in platformization. The list is not decorative. It is a checklist for infrastructure that wants to mediate public life.

Platforms are often defended in the language of convenience, innovation, participation, and user choice. The Platform Society asks what happens when those values crowd out others that democratic institutions are supposed to protect. A transport app may increase convenience while weakening labor protections or local planning. An education platform may expand access while creating dependency on proprietary learning analytics. A health platform may make self-tracking easy while turning intimate data into strategic assets. A news platform may distribute journalism widely while making public attention dependent on opaque selection systems.

The conflict is not simply public sector versus private sector. The authors treat markets, governments, and civil society as actors in a contested field. That is more realistic than either platform boosterism or simple anti-corporate complaint. Public values do not defend themselves. They have to be articulated, designed for, regulated, funded, and enforced.

A useful test is whether a platform can prove how a value is protected in the product. Privacy means data limits and deletion, not only a policy. Accuracy means correction paths and source provenance, not only engagement. Safety means risk assessment and incident response, not only trust-and-safety branding. Accessibility means usable participation across disability, language, age, and resources. Democratic control means that affected publics, researchers, regulators, workers, and civil society can see enough evidence to contest the platform's account of itself.

That proof should be attached to remedies. A user should be able to challenge a moderation or ranking decision that materially affects them. A worker should be able to contest a platform score that governs access to income. A school or clinic should be able to export records and change vendors without losing institutional memory. A regulator or qualified researcher should be able to test systemic-risk claims without relying only on press releases or dashboards chosen by the platform.

News, Transport, Health, Education

The sector chapters are the reason the book has aged well. Rather than treating platformization as a single social-media problem, the authors follow its movement across news, urban transport, health care and health research, and education.

Each sector has a different public mission. Journalism is tied to public knowledge and democratic accountability. Transport is tied to streets, labor, safety, and urban planning. Health is tied to care, privacy, scientific research, and unequal vulnerability. Education is tied to development, autonomy, curriculum, public funding, and institutional trust. Platformization changes each sector differently, but the pattern repeats: a public function becomes dependent on data-rich intermediaries whose business models and governance structures are not identical with the sector's public purpose.

This is the book's strongest corrective to generic technology talk. A platform mechanism is never just technical. It enters a domain with histories, laws, duties, professional norms, inequalities, and public expectations. The same dashboard logic can mean different things in a newsroom, a classroom, a clinic, or a city street.

That sectoral reading matters for AI procurement. A summarizer in a newsroom raises source integrity and editorial independence questions. A routing tool in urban transport raises safety, accessibility, labor, and municipal-planning questions. A health assistant raises confidentiality, evidence quality, clinical responsibility, and unequal vulnerability questions. A tutor or learning platform raises child-safety, curriculum, autonomy, disability access, and student-record questions. The governance unit is not "AI" in the abstract. It is the platformized sector and the public mission that can be displaced.

The AI-Age Reading

The book was published before generative AI became the dominant interface story, but its analysis now looks like a prehistory of model platforms. AI systems depend on the same infrastructure that platform society built: data capture, cloud concentration, identity layers, ranking systems, content moderation, developer ecosystems, app stores, payment systems, metrics, and institutional dashboards.

Van Dijck's 2024 commentary "Governing platforms and societies" makes that connection explicit. Looking back on the 2018 book, she argues that platformization has affected labor, business management, democratic processes, and institutions, and that generative AI is consolidating the power of firms that own compute, cloud infrastructure, proprietary data, and AI services. The problem is not only that AI companies may make mistakes. It is that they may become the layer through which other institutions must act.

A 2023 Nature Machine Intelligence comment by Fabian Ferrari, van Dijck, and Antal van den Bosch sharpens the issue for foundation models. They argue that training procedures for systems such as GPT-4 need to be accessible to regulators and researchers, and that openness and publicness are not the same thing. That distinction matters. A model can be nominally open while still depending on private infrastructure, private benchmarks, private distribution, private cloud contracts, and private decisions about whose knowledge becomes machine-readable.

The AI-era extension of The Platform Society is straightforward: when cognition is offered as a service, public values have to be built into the service layer. Otherwise the interface becomes a privatized public square, classroom, clinic, library, workplace assistant, search engine, and administrative front desk all at once.

Model platforms add new governance surfaces: model APIs, hosted weights, fine-tuning tools, retrieval stores, evaluation suites, content filters, agent tools, app stores, payment rails, enterprise connectors, logging defaults, safety policies, and model routers. Those surfaces decide which knowledge is available, which actions are permitted, which actors are discoverable, which harms are recorded, and which users can appeal. This is why AI governance has to include data provenance, audit trails, audits and assurance, AI search, agents, and vendor governance.

The model-platform version of selection is especially easy to miss because it often appears as a single answer. The service has already selected sources, ranked documents, filtered unsafe content, chosen tools, priced tokens, logged interactions, and applied policy before the user sees a sentence. Public values have to reach that hidden chain, not only the final output.

Governance and Safety

The practical governance object is a public-value register. For each platformized function, an institution should record the public mission, affected groups, data flows, ranking or model logic, vendor dependencies, moderation or enforcement rules, appeal path, audit evidence, incident process, and exit plan. Without that register, public values remain vocabulary rather than control.

The register should be versioned like infrastructure. It should identify the deployed service, system owner, data categories, model or recommender components, ads or monetization hooks, third-party processors, retention schedule, human review path, escalation authority, public-contact point, and the trigger conditions for suspension or decommissioning. A platform that cannot produce those fields for a high-impact use is not ready to mediate that public function.

Strong governance begins before adoption. Procurement should ask whether the platform or AI service can export data in usable form, separate public-service data from product improvement, support privacy-preserving analytics, publish meaningful transparency reports, preserve logs for appeal, allow independent audit where warranted, and avoid locking a public function into a single vendor. If those answers are missing, the service is not infrastructure-ready.

Safety also requires proportionality. Child-safety, fraud, harassment, election-integrity, health, and public-security risks are real; platforms need reporting, moderation, rate limits, provenance, and crisis response. But safety controls can themselves harm public values when they expand surveillance, mandatory identity checks, automated over-removal, state pressure, or inaccessible appeals. The goal is not maximum control. It is bounded control with evidence, contestability, data minimization, and independent review.

For model platforms, the safety case must cover distribution rather than only model behavior. A model output becomes socially consequential when a platform ranks it, recommends it, sells ads around it, routes a user through it, stores it in memory, lets an agent act on it, or inserts it into a public-sector workflow. The review record should identify the deployed service, model or system version, data sources, permissions, affected public values, human oversight, appeal route, monitoring plan, and shutdown authority.

That is where current law and standards help, but only if their limits are clear. The DSA turns platform power into more records; the DMA targets gatekeeper bottlenecks; the AI Act attaches duties to some AI providers and deployers; NIST provides voluntary risk-management language; the FTC report supplies evidence about surveillance incentives. None of these alone guarantees a responsible platform society. Together they show what the book already argued: the interface is not outside governance.

The Public-Value Register

The operational artifact is a public-value register for every platformized public function. It is not a marketing transparency page. It is the record that lets affected people, staff, auditors, regulators, and future maintainers reconstruct how a public value was protected or traded away.

That register should connect to the same infrastructure as public registers, vendor review, AI system inventories, audit trails, incident reporting, and the dependency and exit protocol. Otherwise "public values" remain a statement of intent rather than a maintainable control surface.

Where the Book Needs Friction

The book's main limitation is that "public values" can sound more stable than they are. Privacy, security, fairness, accessibility, innovation, democratic control, and accountability can conflict with one another. They also mean different things to workers, patients, students, journalists, municipalities, regulators, platform firms, and users. A governance framework has to decide how conflicts are resolved, not merely name the values in tension.

Those conflicts are not a reason to abandon the vocabulary. They are a reason to demand procedure. A platform society needs institutions that can say why child safety justified one form of friction but not another, why fraud prevention required a signal but not indefinite retention, why a recommendation system optimized one public outcome rather than another, and who gets to revisit the tradeoff when harm appears.

Some reviewers have made a related point. The Romanian Journal of Communication and Public Relations review praises the book's comprehensive treatment of platform-driven social change, while noting its focus on how public values and sectoral activities are shaped by platforms and shared among corporations, governments, and civil society. The useful question is whether that tripartite responsibility is enough when one actor controls the infrastructure on which the others increasingly depend.

The International Journal of Communication review also treats the book as an important intervention in platform studies. The remaining challenge is operational: moving from diagnosis to institutions that can actually inspect, contest, fund, build, and govern alternatives. Public values need procurement rules, audit rights, data-access regimes, labor protections, public-interest technology capacity, appeal systems, and credible exits from dominant platforms.

The book also predates the DSA, DMA, EU AI Act, current Online Safety Act implementation in the United Kingdom, large-scale DSA transparency databases, general-purpose AI obligations, and the present model-platform market. Those developments do not invalidate the book. They make its vocabulary more actionable and reveal its next task: turning public values into enforceable evidence, remedy, and alternative infrastructure.

What This Changes

The lasting lesson is that an interface can carry a constitution without admitting that it is political. A button can encode a market. A feed can encode editorial power. A recommender can encode a theory of relevance. A reputation score can encode labor discipline. A dashboard can encode managerial priorities. An API can encode sovereignty.

The Platform Society gives a disciplined way to read those encodings. Ask what has been datafied, what has been commodified, what is being selected, which public values are being displaced, and who has the power to change the rules. Then ask the same questions again after AI has been added to the system.

The practical warning is simple. Model platforms will present themselves as assistants, copilots, agents, tutors, doctors' helpers, customer-service representatives, coding partners, research tools, and civic interfaces. The real question is what public functions they absorb, what dependencies they create, what records they keep, what futures they rank, and whether democratic institutions can still govern the world once it has been routed through them.

The recurring pattern is not mystical. It is institutional displacement: a public function moves into a platform; the platform turns the function into data and rules; the rules reshape behavior; the reshaped behavior becomes evidence that the platform is necessary. The antidote is not nostalgia for pre-platform institutions. It is public-value engineering: build, buy, regulate, audit, and exit systems according to the missions they now mediate.

The operational standard is simple to state and hard to meet: if a platform shapes public life, it must leave enough evidence for the public to know what happened, enough remedy for affected people to contest it, and enough institutional capacity for society to choose a different arrangement.

Source Discipline

This article separates book facts, conceptual interpretation, regulatory context, standards guidance, and platform-accountability evidence. Oxford Academic and Utrecht University support the bibliographic facts and publisher framing. Academic reviews and van Dijck's later commentary support the interpretation of platformization and public values. EUR-Lex, European Commission, European Centre for Algorithmic Transparency, FTC, Ofcom, EU AI Act Service Desk, and NIST sources support the current governance context checked on June 25, 2026.

Legal claims are jurisdiction-specific and procedural. The DSA, DMA, Online Safety Act, and EU AI Act do not apply to every platform in the same way. A designation, guidance page, investigation, risk report, staff report, or voluntary framework is not the same as a court judgment or final enforcement finding. Current regulatory pages and databases can change after review, so the review date matters.

Transparency artifacts also have limits. A DSA statement of reasons is evidence that a provider reported a moderation decision under a schema; it is not a complete measure of harm, error, user experience, or discrimination. A platform transparency report is evidence of what the platform measured and chose to disclose. Independent audit, researcher access, user testimony, regulator files, and reproducible methods are needed to test those accounts.

This review does not claim that any AI system is conscious, divine, or AGI. It treats AI systems as institutional and technical systems that route attention, evidence, decision support, and delegated action. The governance question is what the platform can make visible, profitable, authoritative, automated, appealable, or irreversible.

Sources

Book links are paid affiliate links. As an Amazon Associate I earn from qualifying purchases.

Buy on Amazon Browse Books

Return to Blog · Return to Books