Maintenance usually remains outside the photograph: it cuts no ribbon, promises no revolution, and rarely produces a headline. Yet it determines who can keep using a tool after the novelty fades, who receives help when something breaks, and who pays with time for decisions that nobody reviewed. In a life shaped by screens, maintenance is a quiet form of justice because it preserves access, corrects inequality, and restores the ability to choose. A responsible digital service is measured not only by what it enables on its first day, but by how it responds on an ordinary Tuesday, with a slow connection, a confusing update, or a person whose needs were absent from the ideal case.

Attention also requires maintenance. An interface may begin clearly and later fill with alerts, recommendations, windows, and inherited obligations. Each addition appears minor, but together they alter the relationship between a person and a purpose. Someone who came to check an appointment ends up resolving consent requests; a student opening a school platform searches for an assignment among announcements; someone hoping to speak with a relative receives an endless sequence of prompts. The resulting fatigue is often presented as a lack of individual discipline, although it partly comes from systems that accumulate friction without reserving resources to remove it.

Maintenance does not mean preventing every change. It means observing consequences, repairing what deteriorates, and retiring what no longer serves. It requires budgets, accountable owners, deadlines, and intelligible channels. It also requires listening to those who find failures before dashboards do: older people, users with disabilities, families with few devices, workers handling incidents, and young people who recognize capture tactics from within. Justice appears when their experience stops being treated as anecdote and becomes knowledge for decisions.

From a brilliant launch to everyday care#

Many organizations reward launches because they are visible and measurable. An application is announced, a download figure is celebrated, and the project is considered finished. Maintenance is divided among teams without authority or left to contracts concerned only with technical availability. But a system can be online and still be unusable. It can load quickly while speaking in incomprehensible language; protect a database while requiring repeated information; record activity while hiding the exit. Technical continuity is necessary, but it cannot by itself guarantee a dignified experience.

Many organizations reward launches because they are visible and measurable. An application is announced, a download figure is celebrated, and the project is considered finished.

Everyday care expands the question. It asks not merely whether something works, but for whom, under which conditions, and at what cost. It examines how long a person needs to complete a task, how many interruptions occur, whether an error can be corrected, and which alternative remains available. It also reviews work transferred to households, teachers, administrative staff, or communities. When an update saves minutes for an institution but forces thousands of users to relearn a journey, that distributed time belongs in the assessment.

A mature maintenance policy distinguishes four commitments. Preventive work finds failures before they cause harm; corrective work responds promptly and repairs; adaptive work follows legal, social, and technological change; evolutionary work improves a service without turning the public into a permanent laboratory. All four need an accountable person, reserved resources, and public priority criteria. Otherwise, every incident competes for attention, and those affecting the most visible users tend to win instead of those causing the greatest exclusion.

Attention as infrastructure that wears down#

Attention is not a private and inexhaustible reservoir. It rests on material conditions: rest, predictability, clear language, response times, and freedom from unnecessary alarms. When every service demands availability, introduces variable rewards, or turns any delay into an emergency, that infrastructure wears down. A person must continually reconstruct what they were doing and why. This recovery cost affects learning, work, care, and conversation, and it weighs most heavily on those already living with constant interruption.

Attentive maintenance reviews the entire ecology, not an isolated screen. It counts notifications, duplication, context changes, expired sessions, and messages designed to produce guilt. It observes whether an objective ends clearly or the design automatically opens another chain. It asks which features were created to solve a need and which remain because they generate commercial activity. The decisive metric is not how long somebody was retained, but whether they achieved what they sought without surrendering more attention than necessary.

This perspective avoids blame. Personal habits matter, but they form within particular architectures. It is disproportionate to demand perfect self-control while products test thousands of variations to increase time spent. Responsibility must be distributed: individuals establish boundaries; teams remove manipulative patterns; leaders protect uninterrupted time; public authorities set and supervise obligations; independent researchers test effects. Maintaining attention means ensuring that the reasonable option does not require a continuous struggle.

Case 1: a school repairs digital communication#

In an illustrative case, a secondary school discovers that families receive notices through four channels. The official platform contains grades, email carries circulars, an application sends reminders, and informal groups try to fill gaps. Students check all of them for fear of missing an assignment. The school had assessed each tool separately, and none appeared excessive; together they created permanent vigilance, inequality between households, and disputes over messages arriving at night.

School leaders form a group including students, families, teachers, administrative staff, and disability support. For two weeks it records types of alerts and approximate time, without reading private content. It then defines one primary channel, two delivery windows, and three urgency categories. Assignments must be published with a deadline and context; late changes require justification; the system provides an accessible digest and a downloadable calendar. A telephone option remains for anyone without a stable connection.

The three-month review does more than count logins. It compares submissions, repeated inquiries, reported anxiety, accessibility incidents, and teacher workload. The number of messages falls while the quality of received information improves. A student reports that a screen reader does not announce a new label; the school treats it as a priority failure, publishes a deadline, and provides an alternative. Justice does not lie in selecting a perfect platform, but in sustaining a process able to listen, repair, and account for its actions.

Case 2: a health center preserves a human route#

A health center introduces digital appointments to reduce queues. Online bookings initially rise, but reception staff notice another pattern: older patients arrive with incomplete screenshots, migrants do not understand the categories, and people sharing a phone miss reminders. The adoption indicator celebrates growth while repair work moves to assistants, relatives, and community associations. The system works according to the dashboard, though it does not work for every life.

The center analyzes incidents and redesigns the journey with patients who are paid for participating. It reduces questions, explains every use of data, allows a draft to be saved, and offers professional translation. An appointment can also be obtained by telephone or in person without a penalty or reduced availability. Reminders follow individual preferences and never reveal sensitive information on a locked screen. A team reviews abandonment, waiting time, complaints, and differences by age, language, and disability each month.

When an update breaks compatibility with an older browser, a protocol already exists: public notice, temporary restoration, an equivalent alternative, and a later analysis. Nobody attributes the problem to a patient's incompetence. The organization recognizes that digital dependence creates an obligation of continuity. Maintenance therefore protects the attention of someone arriving with a concern, prevents repeated histories, and preserves the clinical encounter for what matters. Efficiency stops meaning transferred effort and begins to measure genuine relief.

Case 3: a company stops confusing presence with work#

A professional cooperative finds that its teams remain connected for long days while projects advance slowly. Messaging accumulates mentions, documents duplicate decisions, and meetings compensate for incomplete instructions. People with caring duties feel compelled to respond immediately to demonstrate commitment. The organization could offer a productivity workshop, but instead it reviews the attention infrastructure that it created.

For one month, it classifies interruptions, identifies channels without owners, and asks which information is missing when a task begins. It then establishes protected blocks, an emergency rotation, and a single decision record. Every request includes an owner, date, and purpose. Mass alerts are disabled, managers schedule delivery outside working hours, and online status is no longer used as an indicator. The technical team publishes a change calendar and maintains compatible versions throughout a reasonable transition.

Metrics include errors, rework, overtime, perceived quality, and the ability to disconnect. Instant responses decline, but decisions arrive with better documentation and projects regain momentum. A part-time employee discovers that the daily brief is issued after she leaves; its publication time changes. That small adjustment expresses a larger idea: maintenance does not preserve a rule untouched. It tests whom the rule excludes and corrects it with enough authority.

Budget, authority, and institutional memory#

Maintenance fails when everyone can report and nobody can decide. Every service needs clear ownership understood as a duty of care, not possession. The accountable person must be able to pause a feature, allocate resources, communicate deadlines, and offer repair. That person also needs limits: incidents affecting rights should not be closed without independent verification or listening to the groups involved. Accountability combines the power to act with the obligation to explain.

A budget reveals priorities more honestly than a manifesto. It should include support, accessibility, security, training, research, and eventual replacement. Reserving money from the beginning prevents barrier removal from always competing with a new launch. Public and private procurement can require response times, portability, documentation, user testing, and exit plans. A supplier that prevents repair or migration turns dependence into risk and weakens collective autonomy.

Institutional memory also needs maintenance. Teams change and, without records, resolved failures return. Organizations should document why a decision was made, which alternatives were rejected, which harm was monitored, and when a review is due. That record does not need unnecessary personal data. It should let a newcomer understand commitments and outstanding debts. Learning from an incident means changing design, procedure, or contract, rather than filing an apology and awaiting the next repetition.

Measuring care without turning it into surveillance#

Measurement is essential, but not every measure provides care. An organization can count clicks precisely and remain unaware of whether anyone understood. It can accumulate session recordings and increase the risk it claims to reduce. Proportionate evaluation begins with limited questions: task completion, recoverable errors, necessary time, requests for assistance, and differences between groups. It adds voluntary accounts to understand causes and deletes data when it is no longer required.

Indicators must include harm rather than growth alone. Abandonment, lost sleep, interruptions, complaints, accidental purchases, accessibility barriers, and transferred workload are relevant signals. Before launching a feature, the organization defines thresholds requiring correction or a pause. Publishing aggregate results supports learning without exposing individuals. Independent auditing is particularly valuable when a business model rewards retention or a service determines access to education, employment, health, or rights.

Repair itself must also be measured. How long did the organization take to acknowledge the failure? Which alternative did it offer? Did it compensate for harm, explain the cause, and prevent recurrence? A quick response does not erase a repeated pattern. The objective is to reduce both frequency and impact while distributing the learning. Trust does not arise from pretending to be infallible; it grows when conduct after an error is predictable.

A maintenance practice for everyday life#

Families, schools, and teams can adopt a simple quarterly review. First, they list essential tools and the purpose of each. They then identify alerts, duplicate accounts, permissions, barriers, and accountable people. They ask what would happen without a connection, without the usual device, or in the face of an unanticipated need. They choose one concrete improvement, set a date, and test its effect. The review ends by removing what is unnecessary, because maintenance also requires knowing how to close.

At home, the conversation should include adults and children without turning control into spying. They can agree which alerts are useful, which moments require calm, and how to respond to an emergency. At school, students can analyze design patterns and propose alternatives. At work, the right to disconnect must appear in settings, deadlines, and managerial behavior. In every setting, exceptions should be handled discreetly, without forcing anyone to explain a personal circumstance in public.

Maintenance is quiet because its success often resembles normality: a link works, an explanation arrives in time, an alternative remains open, and somebody can finish without being captured by another chain. It is justice because that normality does not distribute itself. It takes curiosity to find deterioration, responsibility to assign authority, and care to repair without blame. When attention returns to its owner, technology makes room after fulfilling its function. That act of withdrawal, repeated day after day, is as valuable as any innovation.

Institutional sources#

  • European Commission, Digital Services Act and platform responsibilities: https://digital-strategy.ec.europa.eu/en/policies/digital-services-act-package
  • World Health Organization, guidelines on physical activity, sedentary behavior, and sleep: https://www.who.int/publications/i/item/9789241550536
  • UNICEF, children and well-being in a digital world: https://www.unicef.org/innocenti/reports/children-digital-world
  • United States Federal Trade Commission, consumer guidance on dark patterns: https://www.ftc.gov/news-events/topics/truth-advertising/dark-patterns