A woman opens a public service site to book a medical appointment. The service asks her to choose a region, create an account, confirm a code and return to a page whose earlier form has disappeared. A bright invitation to download an app remains easy to see. The route to speak with a person is difficult to find. She leaves without booking. The design did not merely affect her experience; it changed what she could do.
Digital environments that strengthen rather than capture agency is the Institute’s design synthesis. It concerns whether an environment lets people understand their options, pursue their own purposes, leave a process and repair a mistake. It is broader than reducing time on a screen. A digital service can consume little time and still block a vital action, or support a long, purposeful conversation that a simple screen-time measure would misclassify.
Start with the person’s purpose
The user arrived to book care. The service may be optimized for account creation, app adoption or fewer calls to a help desk. Those goals can be legitimate, but they are not the same as the user’s. A design review should write down the original task and follow it from arrival to completion. At each step, ask whether the user knows what happens next and whether they can recover from a wrong turn.
Agency is not unlimited choice. A medical booking service needs constraints, security and triage. The question is whether those constraints are explained and navigable. If a route is closed, the person should understand why and what alternative remains. If a field is required, its purpose should be clear. A form can be simple without being opaque.
The public-service example also shows why agency cannot be judged by clicks alone. A person may move quickly through several screens yet never reach a usable appointment. Completion, comprehension, time, error recovery and access to a human route are different outcomes. The Institute’s proposed test keeps them visible together.
Choice architecture shapes action
The UK Competition and Markets Authority describes online choice architecture as the way digital environments present and structure decisions. Its evidence review surveys practices involving choice structure, information and pressure. Some designs help people find a relevant option; others can steer them toward choices they would not otherwise make or make a disfavored option harder to select. The review is a map of evidence and potential harms, not a verdict on every interface with a default.
Defaults can save work when they reflect an ordinary preference and are easy to change. They can also make consent or purchase appear already decided. A countdown may communicate a real deadline or create artificial urgency. An extra confirmation can prevent a costly mistake or trap a person trying to cancel. Ethical assessment needs the purpose, evidence and consequences of the specific design.
The US Federal Trade Commission’s report on dark patterns documents deceptive or unfair practices such as hidden terms and difficult cancellation paths. Its consumer-protection frame is useful for recognizing that visual hierarchy and navigation can have material effects. The Institute’s wider concern includes public services, education and attention, where a design can undermine agency without a purchase occurring.
Attention can be invited or taken
A notification can warn someone about a changed appointment. It can also interrupt them because the platform values return visits. The two may look identical on a device. The difference lies in relevance, timing, control and the user’s own objective. Asking whether a feature increases engagement does not show whether it helps the person accomplish what they came to do.
A design team can make a simple inventory: what prompts appear, why, when and with which opt-out? It can observe whether users can finish a chosen task without unrelated demands. A quiet mode, a clear stopping point and notification settings that work as expected are possible design responses. They need testing with the people and tasks concerned, not only a positive engagement metric.
Attention is not a moral property of a user. Someone who follows a bright prompt may be responding reasonably to the cues provided. Framing a design problem as deficient self-control can leave the architecture unexamined. The Institute’s synthesis is to treat attention as shared between person, interface and institution.
Screen time is a blunt instrument
The OECD’s 2025 brief on screen time and subjective well-being uses cross-country poll data from 14 countries. Such data help describe associations and variation, but they do not establish that a particular amount of screen use causes a particular outcome. Context, activity and prior conditions matter. The brief is a useful prompt to ask better questions about digital experience, not a universal prescription for a daily number of minutes.
Research reviews of social media and well-being have also noted disagreement among pooled findings and limitations in measuring use. “Active” and “passive” categories may conceal many different experiences. A person can actively argue in a hostile thread or passively enjoy a friend’s photograph. The measured category alone cannot tell us whether the activity supported agency or well-being.
For service design, the relevant measure may be successful task completion, clarity and ability to choose. A long call with a clinician may be desirable. Repeated attempts to find an appointment are not. Reducing both to screen time would reward the wrong change.
What a Swedish report reveals
Internetstiftelsen’s Svenskarna och internet 2025 reports that close to one in four adults in its survey had refrained from booking a doctor’s appointment because the e-services were too difficult. The report’s underlying Swedish panel was sampled and weighted, but it remains self-report about experience and behavior. The result does not establish that every missed appointment was caused by interface design alone. It does make the consequences of usability concrete enough to investigate.
The same report notes that many people ask AI tools questions instead of using a search engine. A fluent answer can make the path from claim to source less obvious. The Institute’s response is not to ban a useful route. It is to ensure that people can inspect a consequential claim, find original information and recognize when a tool has synthesized rather than retrieved it.
Nordic digitalization is often described through high rates of internet use. High access is important, but it does not prove that every essential service can be completed. An inclusive standard should examine those who stop midway, need assistance or never start because they expect the route to fail.
Strengthen the ability to leave
A person should be able to stop, save progress and return without losing context where the service permits it. If they decide not to subscribe, share data or install an app, the exit route should be clear. This does not require eliminating all friction. It requires that friction serve a justified task, rather than one party’s preference for retaining attention or data.
Cancellation is an unusually revealing test. Compare the effort to join with the effort to leave: number of steps, clarity of labels, waiting time and need for a human conversation. A service may need identity checks to prevent fraud, but they should be proportionate and explained. If an exit is technically available but practically hidden, the user’s choice is weakened.
In a public service, “leave” may mean switching channel. A person unable to finish online should have an accessible telephone, desk or supported route when one is necessary to receive an essential service. Providing an alternative is not a failure of digital transformation. It is evidence that the service recognizes different capabilities and circumstances.
Make repair ordinary
People enter a wrong date, lose a code or discover that a record is outdated. A system can treat these as exceptional failures or as normal cases that deserve a clear recovery path. Error messages should name what happened in plain language and preserve valid work when possible. A correction should be visible, and its effect should be explainable.
Repair matters especially when a system’s output affects someone else. The person should know how to contest inaccurate data, what evidence is needed and when a response will arrive. Staff need an escalation route for cases that the interface cannot represent. A “contact us” page that never reaches an empowered reviewer is weaker than it looks.
Testing should include a wrong turn. Give participants a realistic mistake and watch how they recover. Record whether they find help, whether progress is saved and whether they understand the final state. A perfect-path demonstration can miss the part of the service where agency is most fragile.
Measure the burden of a task
Success rates should be broken down by relevant groups and circumstances, while respecting privacy. A service may perform well in a general test yet fail for a person using a screen reader, a small device or a language they know less well. A design can also fail when someone has little time, is worried or must complete a task on behalf of another person.
Time-to-completion can help, but it can mislead. A user who pauses to read a meaningful explanation may take longer and make a better decision. Another may finish fast without understanding the consequences. Evaluation should pair time with comprehension, errors, confidence in next steps and the option to ask for help. The aim is usable choice, not merely speed.
Organizations should examine abandonment carefully. A person may leave because their need changed, because they found another route or because the service defeated them. A raw abandonment number cannot distinguish these possibilities. Short interviews, support logs and task observation can help, provided participation is voluntary and data practices are proportionate.
Design incentives belong in the review
A platform paid by advertising may have reasons to keep people engaged. A public agency may be pressured to reduce staffing costs. A school may want a dashboard that makes participation easy to count. These institutional goals shape interface choices. An agency audit should name the goals and ask where they conflict with the user’s task.
A team might state a design rule: no prompt should interrupt an essential task unless it is relevant to completing or protecting that task. It could record exceptions and test them. Another rule might require equal visibility for accepting and declining a nonessential data use. These are Institute proposals, not claims that one checklist solves all manipulation.
Governance also needs someone empowered to act on results. If tests show that a group cannot book care, the team should be able to change the form, channel or staffing. An excellent usability report with no route to implementation does little for the people who encountered the barrier.
How to test a stronger environment
Choose a representative set of real tasks, including at least one involving an error, a change of mind and an accessibility need. Ask participants what they intend before they start, then observe what they can complete. Afterward, ask them to explain the options they chose and what they think will happen next. The evaluation should compare the interface with an alternative route when one exists.
Longer-term checks can detect hidden costs. Do people return because the service is useful, or because a prior attempt failed? Can they change notification settings and keep those choices? Does a correction remain fixed after an update? Longitudinal evidence is more demanding than a launch-day test, but the claim that an environment strengthens agency is itself a claim about use over time.
The Institute’s proposed outcome is not maximal use. It is the ability to accomplish a chosen task, understand important consequences and leave or change course without unreasonable penalty. That outcome can be measured imperfectly, refined with users and compared across versions.
What the evidence does not show
The OECD poll does not yield a universal causal screen-time threshold. The Swedish survey does not isolate interface design as the sole cause of every missed booking. Choice-architecture reviews and dark-pattern reports identify mechanisms and concerns, but they do not prove that every default or reminder is manipulative. The Institute’s agency framework is a design synthesis, not a validated clinical or legal test.
No single completion metric establishes that people acted freely. A user may complete a form because alternatives were obscured or because a necessary service left no real option. Good evaluation considers comprehension, access, reversibility and the institutional purpose of the design alongside completed clicks.
Return to the booking attempt
In a stronger service, the woman can see the steps before starting, save her place and correct a failed code without re-entering every field. If the online route does not work for her, she can find a human alternative. The site explains what data it needs and lets her finish the booking without an unrelated app invitation taking over the path.
The design has supported agency if the person can do the thing she came to do, understand what happened and change course when necessary. That standard asks more of an interface than keeping her attention. It also gives designers a clearer, more human outcome to improve.
Research references
- OECD. (2025). Screen time and subjective well-being: Insights from a few countries worldwide. OECD Policy Insights on Well-being, Inclusion and Equal Opportunity, No. 24. View source
- Internetstiftelsen. (2025). Svenskarna och internet 2025: Press summary and survey methods. View source
- Competition and Markets Authority. (2022). Evidence review of online choice architecture and consumer and competition harm. View source
- Federal Trade Commission. (2022). Bringing dark patterns to light. View source
- Valkenburg, P. M. (2022). Social media use and well-being: What we know and what we need to know. Current Opinion in Psychology, 45, 101294. View source