Cities already feel alive. A responsive city takes this intuition and turns it into action; it uses sensors, data, and intelligence networks to perceive what is happening, remember patterns, and predict what might happen next. This is less about devices and more about learning loops that connect streets, buildings, services, and citizens to make daily life function better.
At its best, urban sensing means the four elements working together: perception, memory, inference, and action. Perception collects signals such as traffic density or air quality. Memory stores these in forms the city can search. Inference turns this historical data into predictions. Action changes signals in the world, such as resetting traffic light timing or altering bus routes. Digital twins like Virtual Singapore demonstrate how these layers can be combined to simulate options before a city takes action. This reduces risk and helps align choices with public objectives.
Sensitivity always brings governance issues with it. Who owns the data, who is responsible for the decisions recommended by the algorithms, and how are citizens involved in this process? Real projects have revealed both promising and problematic aspects. Therefore, when evaluating proposals, definitions that emphasize foresight as well as automation are useful.
The Conceptual Foundations of Sensitive Architecture
Defining the concept of “sensitivity” in urban form
A responsive city is one that can perceive, relate, and anticipate. In other words, it remembers what has happened, relates these memories to the present, and uses them to decide what to do next. This framework, commonly used in architectural discourse, distinguishes responsiveness from mere connectivity. It focuses on the experience within the city as much as it does on the infrastructure.
Key concepts
- Perception and memory: Urban systems continuously collect signals and store them in a way that is useful not only for display panels but also for learning.
- Forecasting and feedback: models predict possible futures, then policies or control systems take action and learn from the results.
- Materialization: Sensitivity manifests itself in built forms as well as in software; for example, in responsive facades or street networks that reorganize traffic.
Real-world applications
- City-scale digital twins: Virtual Singapore enables organizations to test scenarios such as floods, mobility, or construction phases before implementation by integrating detailed 3D models with real-time and historical data. This is a combination of perception, memory, and inference in a single tool.
- Learning facades: Abu Dhabi’s Al Bahar Towers use a dynamic mashrabiya that opens and closes to manage exposure to sunlight throughout the day, converting sensor inputs into physical changes on the building’s exterior.
Historical precedents and parallels
The roots of responsive architecture lie in the cybernetic science of the mid-20th century. Cedric Price and theater director Joan Littlewood’s work, Fun Palace, envisioned a cultural machine that would constantly reconfigure itself according to use. Price, working with cyberneticist Gordon Pask, developed control systems and feedback mechanisms designed to enable architecture to learn from people. Although never built, this project laid the groundwork for subsequent experiments linking information, behavior, and spatial change.
In the 1970s, Price’s Generator project, conducted in collaboration with John Frazer, proposed an environment that could reconfigure itself according to user preferences. This was an early example of an “intelligent” building, conceived as a system rather than a static object. During the same period, Nicholas Negroponte’s Architecture Machine Group at MIT explored the possibilities of computers collaborating with humans in design and documented methods for adaptable environments. These studies positioned architecture not as a one-time artifact, but as an ongoing dialogue between users, computation, and space.
https://eliza-pert.medium.com/1976-22121bb498c4
Lessons to be learned from their careers:
- First, feedback: the most effective projects focused on cycles of perception, learning, and restructuring.
- Insiders: The goal was not full automation, but to enhance human influence by enabling environments to respond to public use.
- From theory to practice: Today’s dynamic facades and urban twins are the heirs to these speculative prototypes.
Differentiating from “smart” and “sensitive” cities
Smart city initiatives typically focus on using digital technologies to make services more efficient in areas such as transportation, energy, water, and administration. Responsive design generally refers to buildings or areas that respond to stimuli in real time, such as a facade that opens with the sun or a plaza whose lighting changes depending on usage. Responsive architecture goes even further by emphasizing the long-term memory, foresight, and management of these learning systems.
Practical differentiation
- Smart: digitize and optimize services. Example: open data portals for coordinated traffic management or city services.
- Responsive: Immediate response to inputs. Example: Al Bahar Towers’ kinetic shading system that reduces heat gain by tracking the sun’s position.
- Responsive: memory plus foresight plus controlled action. Example: Virtual Singapore’s ability to simulate policies before implementation, along with inter-agency governance.
The importance of this distinction
- Social license and trust: Toronto’s Quayside project faced some challenges due to the inability to find a convincing solution regarding data rights and public oversight. This situation demonstrated that sophisticated technology can come to a standstill without an established governance structure.
- Urban vitality: Songdo’s experience reminds designers that technical integration does not guarantee vitality. Learning systems should value adaptability, public life, and local intelligence as well as efficiency.
Design proposal
Aim for systems that learn not only about people, but with people. Use clear public value statements, transparent data policies, and feedback loops that residents can see and shape. This is the practical essence of a responsive city.
Theoretical Foundations and Philosophical Frameworks
Cities are not merely physical. They are cognitive, sensory, and cultural systems that help people think, move, and create meaning. Responsive architecture draws on these traditions to answer a simple question: How can a space feel, be remembered, and be navigated while keeping people in control?
Key concepts you will see as references
- Ecological perception and possibilities: Spaces invite certain actions. Stairs, benches, passages, and thresholds show you what you can do before any sign tells you.
- Extended and distributed mind: tools, signs, streets, and digital systems can become part of our way of thinking. The city is a socio-cognitive skeleton.
- Ubiquitous computing and sensing: technology should “disappear” from daily life and allow environments to quietly support attention and activity.
Where theory meets practice
- Spatial syntax treats street layouts and building plans as information structures that shape movement and social life. It yields tangible results in stations, hospitals, and city centers.
- Urban digital twins, like Virtual Singapore, connect perception, memory, and simulation, enabling organizations and communities to test options before taking action.
Cognitive and sensor-based space philosophy
1) Ecological perception: affordances and design
James J. Gibson’s concept of affordances states that people perceive their action possibilities directly in their environment. Good urban design makes these possibilities understandable without extra signage. Tactile curbs that say “slow down,” curb edges that invite safe turns, and benches that invite sitting are simple examples. Research and practice demonstrate how affordances can inform architectural theory and everyday design decisions.
Real-world lens: debates about hostile design prove this point. Sharp edges, segmented benches, or the Camden bench “enable” exclusion by design. If a space can be inviting, it can also be deterrent; this has social costs.olan bir tasarım tercihi.
2) The city as an extended mind
Cognition does not end within the skull. Andy Clark and others argue that tools, notes, routes, and devices have become part of thinking. Recent studies link these theories to planning, defining cities as distributed socio-cognitive architectures that guide memory, attention, and cooperation. In practice, this is seen in street layout and sign-based route finding or in everyday phone-assisted navigation that integrates minds with maps.
Real-world lens: spatial syntax functionalizes this by measuring how patterns affect movement. Research uses this to predict pedestrian flow and improve wayfinding in complex buildings such as hospitals and stations.
3) Ubiquitous sensing as a cognitive layer
Mark Weiser’s vision of ubiquitous computing calls for technology to step back. Mark Shepard’s work, Sentient City, extends this to the urban scale, enabling sensors, mobile media, and networks to “remember, relate, and anticipate” places. Today, urban digital twins put this idea into practice by simulating traffic, flood risk, or microclimate options before anything is built.
Phenomenology, perception, and impact in architecture
1) Body first: the sensations evoked by spaces
Phenomenology focuses on lived experiences. Merleau-Ponty emphasizes perception not only through sight but also through the body. Juhani Pallasmaa applies this to architecture and advocates for multi-sensory design that values material, texture, sound, temperature, and smell. Christian Norberg-Schulz adds the concept of genius loci: architecture should support the spirit of the place and provide people with an existential foundation.
Design actions you can take: choose materials that age well, adjust acoustics and lighting for comfort, and arrange thresholds, arrivals, and directions so they feel natural rather than forced. These thinkers’ reviews and summaries consistently emphasize the shift from eye-centered design to sensory design.
2) Agency through spatial configuration
When we examine how the selection of arrangements is made possible, phenomenology meets analytics. Spatial syntax shows that integration and visibility are related to the places where people actually move. Designers use this information to improve wayfinding, reduce stress, and support social interaction. For example, health studies compare loop corridors with tree-like corridors to reduce confusion and save time.
Design lesson: Combine the feel of the space with movement logic. Map out visual areas, shorten paths to important destinations, and support safe navigation by combining clear sightlines with sensory cues such as sound and temperature.
3) The meaning and memory of space
Phenomenology reminds us that spaces carry stories. Streets and courtyards hold collective memory. When we add perception and data, we should amplify these narratives rather than flatten them into dashboards. Critics such as Shannon Mattern warn that treating a city like a computer means overlooking human knowledge already embedded in institutions like libraries and daily routines.
The ethics of embedding intelligence in the built environment
1) What could go wrong?
Smart environments can drift toward surveillance or biased automation. Predictive policing has faced intense criticism for reinforcing historical biases, and some systems have demonstrated low accuracy rates in practice. Computer vision tools used on streets and in vehicles have documented biases, such as higher miss rates for children in pedestrian detection. These risks are not just software issues but concerns related to urban design.
A cautionary example regarding governance: Sidewalk Toronto proposed a civic data trust to manage “urban data,” but the plan struggled to gain public trust and was ultimately canceled. This situation highlighted how the uncertainty surrounding data rights can hinder ambitious projects.
2) What emerging railing designers need to know
- EU Artificial Intelligence Act: Effective from August 2024, it includes phased obligations from 2026 onwards. It prohibits certain uses, regulates high-risk systems, and adds literacy and governance rules concerning urban artificial intelligence.
- NIST AI Risk Management Framework: a voluntary standard organized around four functions, it is widely used to structure the management, mapping, measurement, and control of AI risks.
- OECD AI Principles: international guidelines for democratic, trustworthy, human-centered artificial intelligence.
- UN-Habitat People-Centered Smart Cities: non-binding guidelines that promote equity, participation, and capacity building in urban technology strategies.
3) Responsible design guidelines
- Minimize and localize data: collect only what you need, keep it close to where it’s used, and keep the default retention period short.
- Make the models visible: Publish what the system detects, how decisions are made, and how results can be challenged. Use public displays and online logs that citizens can read.
- Co-governing with communities: Establishing committees that include residents, disability rights advocates, small businesses, and civil society to review usage examples and audits. Lessons learned from Toronto demonstrate that governance must be supported not only with words but also with concrete actions.
- Conduct bias testing on important issues: If computer vision affects lighting, intersections, or policing, monitor its performance in terms of age, gender, and race. Research shows why this is important for safety and fairness.
- Choose human control: Keep ultimate authority with accountable public officials and combine automation with explicit manual override procedures according to standards such as NIST AI RMF.
Components and Technologies of the Sensitive City
Sensor networks, IoT, and data collection
What you need to know
Cities perceive the world in many ways: air quality nodes, traffic detectors, people counters, weather stations, structural strain gauges, water and waste monitoring devices. Two open families are invaluable for enabling these heterogeneous devices to speak a common language: the OGC SensorThings API (for defining devices, observations, and even sending tasks to actuators) and OPC UA (for secure, vendor-independent data exchange in buildings and industry). These standards ensure that data is structured, queryable, and interoperable between different vendors..
The connection layers you will encounter
- LPWAN for Urban Areas: LoRaWAN provides battery-friendly sensing at a kilometer scale for air, water, parking, and flood networks. NB-IoT offers a similar range but operates on licensed cellular spectrum with carrier-class service levels. Both target small, infrequent messages from hard-to-reach locations.
- Message transfer: MQTT is a lightweight publish-subscribe backbone that transfers reads from devices to brokers and applications with minimal bandwidth. It is an OASIS standard widely used in IoT and IIoT.
- Short-range and site systems: Within buildings and facilities, Wi-Fi, BLE, and OPC UA gateways collect sensor data and securely deliver it to higher-level services. OPC UA is standardized as IEC 62541 and supports both client-server and publish-subscribe models.
Real-world applications
- EU smart street lighting pilot projects combine motion sensors and time-based dimming to enhance security while saving energy, demonstrating how LPWAN and MQTT can contribute to urban lighting platforms. Studies highlight adaptive lighting architectures and verified savings pathways.
- Digital twins such as Virtual Singapore take sensor streams and test city-scale flood, mobility, and construction scenarios, transforming raw observations into shared operational awareness.
Data processing, artificial intelligence, and feedback loops
Edge to cloud
Processing is a continuous process. Edge devices and gateways filter and aggregate data close to the source to enable rapid response. Fog or MEC nodes aggregate across regions. Cloud platforms store history and train models. This layered approach reduces latency and bandwidth while keeping intensive analytical processing where it is most efficient.
Flow and learning
- Pipelines: MQTT brokers transfer data to streaming platforms such as Apache Kafka, enabling applications to subscribe to real-time topics and combine them with historical data for analysis and alerts.
- Edge models: Compact runtimes such as ONNX Runtime distribute trained models to cameras, controllers, and kiosks for on-device inference, improving privacy and response times.
Closed loops in practice
- Traffic signals: The SURTRAC project in Pittsburgh adjusts lights in real time using decentralized, AI-based optimization. Field results and transportation summaries reported by CMU show significant reductions in travel time and delays in pilot corridors.
- HVAC control: ASHRAE Guide 36 enables building controllers to efficiently maintain comfort and eliminate surface errors by coding high-performance sequences. In addition to these sequences, model predictive control and reinforcement learning studies have reported double-digit energy savings in field tests and provided increasing guidance on how to implement them reliably.
Lessons to be learned from the design
Create visible and testable feedback loops: detect, predict, act, and show the public what has changed. Use edge filtering to protect privacy, open standards for interoperability, and reference sequences to ensure systems continue to function smoothly when AI is offline.
Actuators, dynamic systems, and morphing elements
Adaptive coatings and daylight control
Responsive envelopes convert data into geometry. Electrochromic glass changes color to reduce glare and solar heat without using blinds, thanks to ion movement controlled by predictive controllers and sky sensors. Al Bahar Towers’ kinetic facades, like mashrabiya, open and close to track the sun. Jean Nouvel’s Institut du Monde Arabe pioneered camera-like diaphragms that modulate light at the aperture.
Actuators at the room and city level
- Public space systems: Adaptive street lights brighten when people are present and dim when empty; verified architectures and case studies exist across Europe. Variable speed limits on smart highways use bridge signs to smooth traffic flow and enhance safety when conditions change.
- Transforming buildings: The Shed in New York rolls out a telescopic ETFE-covered shell on rails, transforming a plaza into a climate-controlled hall in minutes. This is a clear example of data-driven action at an architectural scale.
The interconnectedness of everything
Task interfaces in standards such as OGC SensorThings enable software to convert analytical decisions into actionable commands, allowing the same data model that stores sensor readings to also control devices that operate blinds, valves, fans, lights, or facade panels. This closes the loop between sensing and acting in a traceable and controllable manner.
Application checklist
- Interoperable descriptions (SensorThings, OPC UA) paired with lightweight messaging (MQTT).
- First, select the coverage: Open, LoRaWAN for low-speed devices across the city, or NB-IoT for licensed cellular reliability.
- For security and latency, keep the edge logic close to the actuators, then synchronize the summaries with the cloud for learning and control.
Spatial Morphologies and Design Strategies
Adaptive facades and responsive cladding systems
What are they and why are they important?
Adaptive facades change their optical or geometric state to balance daylight, heat, and glare throughout the seasons. Recent studies show that dynamic facades can reduce energy consumption while increasing comfort when properly controlled and integrated with building systems. For example, electrochromic glass has been documented to offer savings potential and health benefits when used with smart controls.
The types you will encounter
- Kinetic shading systems use articulated elements that open and close with the sun. The Al Bahar Towers in Abu Dhabi can be cited as an example, where mashrabiya modules respond to the sun’s position and wind conditions.
- Electrochromic and thermochromic glass changes color to limit heat gain and glare while preserving the view. Meta-analyses show a significant reduction in cooling and lighting needs.
- Double-walled facades create a ventilated cavity that can be monitored and operated with an increasing number of IoT sensors for climate control and maintenance.
How to design for performance
Start with climate and usage. Set clear targets for daylight autonomy, glare, and maximum cooling, then select a facade mechanism that can truly meet them. Use simulations calibrated for annual and extreme conditions, and link the facade logic to HVAC and lighting via a controller. Field-proven examples such as the camera-like diaphragms of the Institut du Monde Arabe and the kinetic curtain of Al Bahar Towers demonstrate both the promise and necessity of robust maintenance strategies.
Flexible interiors and multi-purpose spaces
Principles of adaptability
Be prepared for change by separating long-lasting structures and services from short-lived partitions and equipment. Known as the open building concept, this idea allows interior spaces to evolve without major interventions and reduces waste through demountable design. Reviews and manifestos on Habraken’s work summarize the support-filling distinction, which also resonates in today’s DfD research.
Real-world game book
- Spacious, service-rich shells: The Shed in New York transforms a plaza into a climate-controlled hall using a telescopic ETFE-covered shell on rails. This hall features ceiling fixtures and environmental systems designed for rapid reconfiguration.
- Movable and demountable partitions: Gallery and education projects support new programs by reconfiguring rooms overnight using certified modular wall systems, without construction dust or permits.
- Preparing floors and ceilings for the future: Industry guidelines for office equipment recommend raised access floors and accessible ceiling areas to enable the relocation of power, data, and air in line with new regulations. Government workplace standards treat MER and SER rooms as critical infrastructure linked to these pathways.
Design checklist
Select column grids that accommodate multiple furniture layouts and provide ample power and data in the surrounding area and on the floor, and standardize interface heights and panel sizes so that elements can be replaced without requiring custom work. Test rearrangements in the digital twin before creating the initial plan.
Intermediate infrastructures and hidden intelligence layers
What lies between the floor and the void?
Smart buildings rely on layers that are typically invisible to users: raised floors and plenums for services, double-walled voids housing sensors and actuators, and communication cabinets or edge nodes performing local analysis. Research shows that DSF cavities benefit from careful modeling and monitoring, while hardware guides explain how underfloor and ceiling areas enable the rapid rerouting of power and data.
Streets as extension spines
Outside the building, multi-purpose poles are becoming small urban platforms. The EU’s Humble Lamppost project highlights street lamps with adaptable LEDs, air quality sensing, EV charging, and communication features. Cities are using open platforms like Barcelona’s Sentilo to manage these devices across departments.
Edge, telecom, and resilience
Low-latency services are typically located in micro data centers installed in buildings or service rooms, while 5G small cells are mounted on existing street furniture to densify coverage. Industry reports from the GSMA and technical bodies explain how these elements integrate with poles, walls, and transportation assets, and why edge nodes are important for real-time sensing and control.
Design moves you can implement tomorrow
- Separate gaps at an early stage: space for size enhancers, service corridors, and future devices, not just for today’s loads. Refer to hardware and workspace standards for minimum depths and access.
- Specify open platforms whenever possible: Following examples such as Sentilo, manage poles, pumps, and panels through systems that support multi-vendor devices and open APIs.
- Perform detection and actuation operations in the same location: close the loop by pairing facade sensors with controllable shutters in the DSF cavity and street sensors with dimmable LEDs or signs on the same pole. Case studies on DSFs and smart lighting demonstrate why proximity increases reliability.
User Experience and Human-City Interaction
Concrete perception, possibilities, and control
How do bodies read the city?
People don’t analyze cities like electronic spreadsheets. We feel their surfaces, light, sound, and temperature, and act accordingly. Architectural phenomenology reminds us that space is understood not only with the eyes, but with the whole body. Juhani Pallasmaa advocates multi-sensory design that treats texture, acoustics, and even smell as first-class design elements.
Affordances 101
James J. Gibson’s concept of affordances explains why the roundness of a curb encourages slower turning, while benches encourage sitting. In urban design, shaping affordances means making desired actions understandable without extra signage. Contemporary interpretations directly relate this theory to architectural and public space applications.
Somut kanıt
Kavşaklardaki dokunsal kaldırımlar basit ama etkili bir örnektir. Research shows that these sidewalks help visually impaired individuals follow a straighter path, shorten their crossing time at intersections, and improve their walking pattern. In situations where these sidewalks are absent, serious harm occurs. Certain cases that have come to light in the UK have compelled authorities to complete the installation of these sidewalks nationwide.
Design movements
- Create the shape possibilities before adding markings: curb edges, slopes, edge textures, railings.
- Providing clear and local control over systems that affect comfort, such as lighting and air conditioning.
- Make the control logic visible: if a junction is waiting for pedestrians counted by sensors, indicate this on the pole. This ensures that the control is both perceptible and reliable. Grounding the above in concrete perception and affordance theory increases trust, readability, and safety.
Personalization, privacy, and consent in space
Personalization with limitations
Cities can customize light levels, transportation information, or routes based on context. However, any system that identifies an individual, even through device signals, falls under data protection regulations. Under GDPR, individuals have rights of access, objection, and portability, and controllers must be transparent about purposes and storage.
What is considered consent on the street
Location analysis in shopping malls or stations is generally based on Wi-Fi probe requests emitted by phones. Regulators consider this personal data and require strong security measures to be taken. Guidance from the UK ICO and Spain’s AEPD clarifies that consent must be informed, freely given, and revocable, and that PECR rules apply when collecting signals from user devices. In practice, this means clear indicators, easy opt-out options, and avoiding bundled or coerced consent.
New rules for artificial intelligence in the public sphere
The EU Artificial Intelligence Act has entered into force with phased obligations from 2026 to 2027. Certain prohibitions and artificial intelligence literacy obligations began to apply in February 2025, while general-purpose model rules began to apply in August 2025. Stricter measures are expected to be taken regarding biometric surveillance and high-risk systems used by cities. Ongoing guidance and discussions in 2025 demonstrate both the implementation timeline and the sector’s resistance. Strict measures must be taken.
Civil data alternatives
Centralized data storage is not always necessary for personalization. Barcelona’s DECODE pilot projects have supported city services while exploring tools that allow people to keep their data private or share it for the public good. These approaches are consistent with privacy proposals designed for smart city identity and access control.
Design movements
- Select processing at the device or endpoint with total outputs as the default.
- Provide clear, on-site notifications and simple opt-out methods.
- Publish human-readable data usage logs.
- Map projects to a risk framework such as NIST’s AI RMF so that tasks and risk mitigation measures are clearly defined from day one.
Feedback, learning, and the co-evolution of the user and the city
Closing the loop
A responsive city learns alongside its residents. Citizen reporting platforms like FixMyStreet demonstrate how daily feedback on issues like potholes or lighting is directly fed into public works workflows. While analyses caution against bias regarding who reports and where, the model demonstrates a practical and scalable loop between perception, action, and verification.
Participatory sensing and platforms
Citizen sensors complement municipal networks by detecting local issues such as speeding or air quality. Case studies and critical academic work summarize the benefits and limitations of this approach. From a governance perspective, open-source platforms such as Decidim enable residents to propose, discuss, and follow up on decisions, including participatory budgeting.
Digital twins for shared understanding
City digital twins can make complex modeling publicly discussable. Helsinki’s 3D city model is used to visualize options, share data transparently, and support planning discussions. Research examines the governance and technical structure of this model in detail. When residents can see possible future scenarios, feedback becomes more concrete and decisions improve.
Design movements
- Consider residents as co-interpreters rather than passive data sources.
- To reduce feedback bias, combine reporting practices with proactive social assistance activities in underrepresented areas.
- Use twin or shared maps to show the recommendations, their anticipated effects, and what changed after implementation.
- Publish your learning cycles: what you measured, what you changed, and what happened next. These habits build trust and make the city feel more like a partner than a platform.
Case Studies, Proposals, and Prototypes
Important experimental projects and installations
Urban-scale learning systems
- Virtual Singapore is a nationwide digital twin used to test flood response, mobility, and construction scenarios before taking action in the real world. By combining high-resolution 3D models with live and historical data, it enables agencies and researchers to simulate policies and share the results with stakeholders. Its management is carried out by the National Research Foundation of Singapore and the Land Authority of Singapore, which prioritize inter-agency coordination over a single vendor platform.
- SURTRAC, Pittsburgh, offers a decentralized, AI-based traffic signal system that calculates local timing and shares short-term predictions with neighboring intersections. Field results and transportation summaries reported by CMU show that travel time in pilot corridors has decreased by approximately 25%, the number of stops has decreased, and emissions have fallen.
Building envelopes as a tool
- Institut du Monde Arabe, Paris gün ışığını modüle etmek için kamera benzeri diyaframlar kullanır. Güney cephesinin metalik brise-soleil, kültürel sembolizmi sibernetik kontrol ile birleştirmiştir ve on yıllar sonra hala vaka incelemelerinde incelenen kanonik bir duyarlı cephe olmaya devam etmektedir.
- Al Bahar Towers, Abu Dhabi, güneşi takip etmek için açılıp kapanan kinetik bir mashrabiya kullanarak ısı kazanımını azaltırken manzarayı koruyor. Belgelenmiş vaka çalışmaları ve uygulayıcı raporları, parametrik tasarımını ve çalışma mantığını ayrıntılı olarak anlatıyor.
Critical art and public participation
- Mark Shepard’s work, Sentient City Survival Kit, treats ubiquitous computers not merely as something to be installed, but as a subject to be questioned. The speculative devices in this toolkit invite the public to consider issues of surveillance, prediction, and autonomy in streets equipped with sensors.
Kinetic architecture at the room and city scale
- The Shed, New York creates a climate-controlled hall in minutes by placing a telescopic ETFE-covered shell over a plaza, demonstrating how actuation, control, and the building’s facade panels can be scaled to entire buildings.
Unrealized visions and speculative proposals
Cybernetic adaptability
- Cedric Price’s Fun Palace envisioned a continuously reconfigurable cultural machine where feedback and participation would shape the program and form in real time. Academic studies position this work at the intersection of cybernetics, theater, and social design. Generator, by John and Julia Frazer, expanded this logic with computational recommendations and proposed an environment that could reconfigure itself in response to users.
Megastructural urban futures
- Archigram’s Plug-in City project proposed a service mega-structure where components could be changed as needs evolved. By radically separating long-lasting infrastructure from short-lived modules, this project continues to inspire flexible urbanism to this day. Museums and archives preserve these drawings as cornerstones of design imagination.
- Constant’s New Babylon envisioned a planetary network of elevated, reconfigurable sectors for a nomadic society based on play and invention. Major museums continue to exhibit this work as a lens on freedom, technology, and urban form.
Ambition controlled by management
- Sidewalk Toronto (Quayside) developed a comprehensive “smart district” proposal through extensive public consultation, but the project was canceled. While official statements cited economic uncertainty as the reason, independent reports and public records highlight unresolved issues related to data rights and governance. This case has become a reference point for consent, governance, and scope control.
Ongoing live test environments
- Toyota Woven City, located near Mount Fuji, is a neighborhood specifically built to test robotics, autonomous vehicles, hydrogen energy, and new street typologies. Phase 1 construction will be completed by the end of 2024, and the first residents will move in during 2025. This site is positioned as a long-term prototype for urban technology under controlled conditions.
Lessons learned and transferable design insights
1) Create publicly accessible, multi-scale prototypes
Combine early simulations with tangible demos that people can interact with. Digital twins at the country and city scale, such as the 3D city models of Virtual Singapore and Helsinki, demonstrate how scenario testing can reduce the risks of decisions and make trade-offs visible. For exploration at the room and neighborhood scale, tangible interfaces like MIT’s CityScope help non-experts iterate in real time and see the results.
2) Design for maintenance from day one
Kinetic and media facades are impressive, but they require clear O&M budgets, spare parts, and failure modes. The Institut du Monde Arabe’s long-term studies document the wear and tear of mechanical elements and the importance of service access and renewal strategies. Plan for safe manual override and graceful degradation.
3) Separate long-lasting infrastructure from short-lived hardware
Archigram’s plug-and-play logic and Price’s cybernetic plans foreshadowed the concept of open buildings: keep structures and services durable, while keeping interiors and interfaces easily replaceable. This reduces waste and allows buildings and neighborhoods to learn without being completely rebuilt.
4) Put governance on the critical path
The cancellation of Quayside demonstrates that uncertain data rights can halt even well-funded plans. Using human-centered frameworks and recognized principles, define purpose, consent, access, retention, and accountability early on under public oversight. UN-Habitat’s Human-Centered Smart Cities guide and the OECD AI Principles are practical references.
5) Measure what matters for bodies, not networks
Shannon Mattern’s critique reminds us that cities are not computers. Evaluate libraries, clinics, public transportation stops, sidewalks, and parks as information and care infrastructure. Evaluate projects not only by efficiency or sensor counts, but by the experiences they provide.
6) Close the loop and showcase your work
Traffic control pilots like SURTRAC are successful because they link perception, prediction, and action to measurable outcomes that the public can control. Publish inputs, decisions, and outcomes so residents can see and shape how the city learns.
Challenges, the Future, and Critical Thinking
Technical, infrastructure, and scaling limitations
Interoperability, integration of legacy systems, and reliability
Pilots rarely fail sensors. They fail at the points where old and new systems converge. Growing cities do three things early: they adopt common data definitions, they define the information exchange between twins and control systems, and they plan for gradual degradation. Two key standards help: OGC SensorThings for defining observations and handling actuators, and OPC UA for secure, vendor-independent data exchange in buildings and industry. National programs like the UK’s National Digital Twin encourage similar information management frameworks, so independently created twins can communicate securely over time.
City-scale security and operational resilience
As the attack surface expands, CISA and its partners’ advice is clear and straightforward: expect exploitation across supply chains and interconnected systems, and treat secure procurement processes as a top priority from the outset, with zero trust and design. In the EU, NIS2 raises the bar for essential and important organizations, and ENISA’s 2024 Union-level assessment highlights gaps and next steps. Practical checklists are now available for responsible IoT procurement and aligning city technology with NIS2 control families.
Sustainability, IT budgets, and device lifecycles
AI-heavy analytics improve predictions while increasing energy and water usage in data centers. The IEA predicts that, with AI as the strongest driving force, global data center electricity consumption will approximately double by 2030. This situation forces cities to weigh the gains from optimization against upstream power effects. In terms of hardware, research warns that billions of short-lived devices will contribute to carbon emissions and electronic waste unless reuse, repair, and certified recycling are planned into projects. Treat device and computing budgets like other public services: measure, forecast, and disclose to the public.
Socio-political, equality, and governance concerns
Transition periods are real
Europe’s AI Act has entered into force and is being implemented in phases. Prohibitions and AI literacy obligations will apply from February 2, 2025, governance rules and general-purpose AI obligations from August 2, 2025, and most of the highest-risk obligations will apply by 2026-2027. Despite objections from industry groups, the Commission has not changed the schedule. Municipal teams should plan for uses such as biometrics, mobility control, and security systems according to these dates.
Surveillance, bias, and unequal harm
Evidence continues to highlight the risks in predictive policing and computer vision performance. While 2025 human rights reports call for the prohibition of predictive policing practices in the UK, citing discriminatory effects, technical studies document higher error rates for certain pedestrian groups or children when datasets are imbalanced. If detection leads to critical security measures being taken, independent audits and publicly available scorecards are not optional.
Not just performance, but public legitimacy
The Toronto waterfront project shows how uncertainty around data rights and governance can halt ambitious projects, no matter how advanced the technology. The project’s termination and the city’s subsequent data governance efforts offer a lesson for all cities: Before pouring concrete, disclose the purpose, ownership, storage, and accountability to the public.
Possible futures, risks, and design provocations
Edge-first, privacy-preserving learning
In the near future, it will become closer to where analytical data is generated. Federated learning enables multiple sites to train shared models without centralizing raw data. Studies show privacy gains, lower latency, and resilience to connectivity issues. For sensitive areas, this means station- or building-level models that share updates rather than identities.
Open assignment and accountable operation
Closing the loop will transition from controllable, temporary APIs to standard assignment. OGC SensorThings Section 2 specifies how the software can command devices such as shutters, pumps, or drones using the same model that stores observations. Pair this with immutable logs and dashboards that show what was detected, what was decided, and what was acted upon.
Strengthening the stack through regulation and supply chain management
The EU Cybersecurity Act extends security design obligations to a wide range of connected products, with full obligations coming into effect in December 2027. Combined with NIS2 sector obligations, this pushes cities towards secure default settings and better supplier hygiene. Write these dates into contracts and demand third-party evidence instead of promises.
Design provocations to be tested in policy and form
Add sunset clauses to systems that monitor people: authorizations expire unless benefits are re-proven.
Send with a visible manual override feature and publish drills like fire departments do.
Set aside a budget for the end of the devices’ useful life during the purchase process and publish an electronic waste log throughout the city.
Consider energy as a first-class design principle: For any new digital service, explain the expected computational load alongside its social benefits, referencing IEA projections for data center growth.