Blog · Review Essay · Last reviewed June 19, 2026

The Smartness Mandate and Planetary Governance

Orit Halpern and Robert Mitchell's The Smartness Mandate treats "smart" not as a feature label but as a governing style: sense everything, optimize continuously, and call the result resilience. The useful definition is this: smartness becomes a mandate when sensing, prediction, optimization, and adaptation are made the default answer to social, ecological, and institutional failure.

Planetary governance, in this review, does not mean world government. It means a policy grammar in which cities, grids, forests, markets, supply chains, health systems, and climate risks are all made legible through the same sensor-model-optimization loop, then governed as if better feedback were the main form of public capacity.

The Book

The Smartness Mandate was published by MIT Press on January 10, 2023. MIT Press lists the paperback at 336 pages, with ISBN 9780262544511, and describes the book as a critical history of smartness across ubiquitous computing, artificial intelligence, machine learning, resilience, finance, cities, environments, and planetary governance. The Amazon product identifier used on this page matches the ISBN-10 0262544512.

The book grows out of the authors' earlier Grey Room article, coauthored with Bernard Dionysius Geoghegan, on smartness as a political and infrastructural demand. The full book expands that argument across smart cities, ecological management, financialization, resilience, populations, demonstrations, and machine-learning imaginaries. Its central claim is that "smart" names a way of knowing and governing the world, not merely a class of connected devices.

Halpern works in digital cultures and histories of data; Mitchell works across literature, science studies, and biopolitics. That matters because the book does not read smart infrastructure as a gadget category. It reads it as a worldview in which life, environment, market behavior, and public administration become governable by feedback.

Current Context

As of June 19, 2026, the smartness mandate is visible in ordinary governance instruments, not only in vendor slogans. ISO 37122 and ISO 37123 define smart-city and resilient-city indicator frameworks. UN-Habitat's People-Centred Smart Cities programme treats local digital transition as a governance-capacity problem rather than a pure technology purchase. Helsinki's AI Register and the Algorithmic Transparency Standard show the municipal version of the same move: if an automated system helps a city see, route, or answer residents, the public needs a discoverable record of what the system is and how it is governed.

Critical infrastructure is the sharper current test. NIST's April 2026 concept note for an AI Risk Management Framework profile on trustworthy AI in critical infrastructure says operators will increasingly rely on AI across information technology, operational technology, and industrial control systems. The EU AI Act's high-risk categories likewise place critical infrastructure, education, employment, essential services, law enforcement, migration, and justice in the zone where AI use requires stronger records and transparency. The book's argument is therefore no longer only cultural criticism. It is a guide to where procurement, standards, and law are trying to discipline the smartness mandate after the mandate has already entered public infrastructure.

The counter-mandate is not anti-measurement. It is disciplined measurement: name the public purpose, limit the data, preserve non-digital channels, expose vendor dependencies, publish the inventory where lawful, and decide in advance when a smart system must be paused, repaired, or removed.

Smartness as Governance

Halpern and Mitchell are most valuable when they refuse the product brochure. Smart phones, smart grids, smart cities, smart forests, smart logistics, and smart homes are usually sold as upgrades: more sensing, more responsiveness, more efficiency. The book asks what form of politics is hidden inside that upgrade. A smart system does not simply observe a world that was already there. It defines the world as a set of signals to be captured, correlated, simulated, priced, and adjusted.

That makes smartness a close cousin of algorithmic governance. The point is not that every sensor is harmful or every dashboard is false. The point is that "smart" systems relocate authority. Decisions move from public deliberation to optimization loops, from explicit values to performance metrics, from accountable institutions to platforms, vendors, and experimental zones. The system becomes persuasive because it appears adaptive. Its politics can disappear inside the promise that the environment is too complex for ordinary planning.

The chain is simple and powerful: detect, model, compare, optimize, monetize, adapt, repeat. Each step can be useful. Each step also narrows the situation into what the system can sense and what its sponsor can act on. A city can improve traffic response and still make mobility policy less contestable. A building can save energy and still turn tenants or workers into inputs for someone else's operational model. A welfare office can detect fraud and still make ordinary need look like suspicious deviation.

Crisis Without Politics

The book's sharpest target is resilience. In the smartness mandate, crisis becomes permanent background. Climate instability, market turbulence, supply-chain fragility, infrastructure failure, and public-health risk all become reasons to build systems that sense faster and adapt faster. That sounds practical. It can also train institutions to stop asking what kind of future should be built and ask only how quickly the system can absorb the next shock.

Resilience becomes dangerous when it means adaptation without accountability. A flood model, heat-risk index, energy dashboard, or policing forecast may help officials respond to real danger. It can also normalize the distribution of harm: some neighborhoods learn to absorb heat, surveillance, and outage risk while the system celebrates overall responsiveness. The political question is not whether to measure crisis. It is whether measurement becomes a substitute for prevention, repair, redistribution, or refusal.

This is why the book belongs beside the site's reviews of cybernetics, surveillance, automation, and metric rule. Smartness updates old feedback dreams for a world of planetary-scale data capture. The new faith is not that machines are divine or conscious. It is that continuous sensing and optimization can substitute for contested politics. Under that faith, residents become data sources, places become test beds, and failure becomes a reason to install a more comprehensive system.

The Agent Reading

Read in 2026, The Smartness Mandate clarifies AI agents better than many books about agents. An agentic system does not only answer questions. It perceives a situation through available data, chooses tools, and takes action. That is smartness in miniature: the world is rendered as a dynamic field of inputs, affordances, and feedback. The danger is not machine intention. It is delegated action inside an institutional model of what counts as signal, success, exception, and repair.

An agent is a smartness mandate compressed into a workflow. It inherits goals, data access, permissions, tools, logs, evaluation criteria, and escalation rules. A useful agent can schedule care, inspect infrastructure, triage paperwork, or coordinate response. A dangerous one can turn bad policy into fast policy. The right question is therefore not "is the model intelligent?" but "what representation of the world has been authorized to act, and who can inspect, appeal, pause, or reverse that action?"

Governance and Safety

The governance question is not whether a city, grid, school, hospital, or agency may ever use sensors or models. It is whether smartness remains subordinate to public purpose. A smart deployment should publish the service affected, system owner, vendor, data categories, decision force, update cadence, affected population, appeal path, retention rule, evaluation evidence, and shutdown condition. Without those basics, "smart" becomes a procurement adjective for unreviewable power.

The safety boundary should be drawn around action, not branding. A dashboard that only "supports situational awareness" can still change patrol patterns, emergency response, maintenance budgets, building operations, benefit routing, or school interventions. A system that does not make the final decision can still define the problem, rank the queue, label the exception, or make nonuse look irresponsible. Governance has to record how the smart layer changes attention, not only whether it formally automates a decision.

NIST's AI Risk Management Framework is useful because it treats risk as lifecycle practice, organized around govern, map, measure, and manage functions. NIST's own AI RMF page also notes in 2026 that the 1.0 framework is being revised and that a critical-infrastructure profile is in development. That current context matters for smart systems because the risks sit in the whole arrangement: sensors, databases, models, contracts, operators, maintenance, feedback, and downstream action.

The EU AI Act shows why smart infrastructure is not a neutral category. Annex III treats several uses as high-risk when they concern critical infrastructure, education, employment, essential services, law enforcement, migration, and justice. Articles 12 and 13 require, for high-risk systems in scope, logging capabilities and information that lets deployers understand and use the system appropriately. The lesson for any jurisdiction is practical: if a system cannot produce records, explain its intended use, support oversight, and survive challenge, it should not be trusted as public infrastructure.

Smart-city standards such as ISO 37122 for smart-city indicators and ISO 37123 for resilient-city indicators can help governments compare practices and define measurement methods. They do not settle legitimacy. A metric can make quality of life visible while also becoming a procurement target, a ranking device, or a reason to ignore unmeasured harms. The standard gives vocabulary; democratic governance still has to decide what should be measured, by whom, for whom, and with what right to refuse.

Helsinki's public AI Register offers a concrete counter-mandate: make algorithmic systems discoverable before they become ambient facts. A register is not enough by itself, but without an inventory, affected publics cannot know what to challenge. For public-sector AI and smart infrastructure, an inventory should connect to procurement files, audit trails, human oversight duties, data minimization, notice, appeal, and incident reporting.

Where the Book Needs Care

The book is dense, sometimes deliberately so. It rewards readers already comfortable with media theory, cybernetics, biopolitics, infrastructure studies, and critical theory. That density is part of its force, but it also limits its public usefulness. A city official, union organizer, procurement lawyer, or community group may need a plainer map of where smartness enters a contract, a dashboard, a zoning plan, or a workplace rule.

The other limit is strategic. Critique can show how smartness makes itself feel inevitable, but institutions still have to choose. Should a city refuse a predictive platform, require public audits, narrow data collection, mandate appeal rights, or build noncommercial infrastructure? The book gives a strong genealogy of the mandate. It leaves more room for the practical counter-mandate: procurement limits, public registers, data minimization, community review, labor accounting, appeal rights, open standards, and non-digital service channels.

What This Changes

The Smartness Mandate gives this site a useful diagnostic. When a system is called smart, ask what crisis justifies it, what population supplies its data, what metric defines its success, and who can stop it once it is installed. Ask whether the system improves public capacity or merely makes public life more measurable for someone else.

The book's best lesson is that smartness is not intelligence. It is a political arrangement that treats data capture and adaptive control as common sense. AI governance has to begin there, before the dashboard, before the agent, before the model. Audit the mandate before auditing the model: what crisis is being named, who supplies the data, who gets optimized, who can refuse, who can appeal, who can shut the system down, and what non-computational institution is being displaced?

Source Discipline

This review separates three kinds of evidence. Bibliographic claims come from MIT Press, with Google Books used only as a secondary listing. The genealogy of the argument comes from the earlier Grey Room article. Current governance claims come from standards bodies, official AI governance pages, legal text, and public-sector registers.

Those sources have limits. ISO standards document indicators and methods; they do not prove that a smart-city project is democratic or safe. The EU AI Act source pages are jurisdiction-specific, and official service-desk summaries are explanatory rather than legally binding. NIST's AI RMF is voluntary guidance, not a statute. None of these sources supports claims that AI systems are conscious, divine, or AGI. The argument here is about institutions, evidence, and delegated control.

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