When a technology has just arrived, discussion is often led by demonstrations, adoption figures and promises of rapid change. That moment can be useful for exploration, but it is not enough to decide whether a tool deserves a lasting place in a school, home, community group or public service. Responsible evaluation begins after enthusiasm fades: then real uses, sustained benefits, hidden costs and the people left out can be seen. The question is not whether to choose enthusiasm or rejection, but whether to ask for evidence before turning novelty into dependence.

Time turns a promise into a testable question#

A demonstration shows what a product can do under favourable conditions; serious evaluation observes what happens over months with diverse people, imperfect connections, limited budgets and real responsibilities. An educational application may be engaging in its first week while adding repetitive work for teachers, collecting more data than needed or leaving behind people who need accessibility support. A messaging service may make contact easier while normalising constant availability that wears down young people and families.

A demonstration shows what a product can do under favourable conditions; serious evaluation observes what happens over months with diverse people, imperfect connections, limited budgets and real responsibilities.

It is therefore useful to replace absolute claims with specific questions: which defined problem did it solve, which people benefit steadily, which new difficulty has appeared, and what evidence supports the result? The tool should also be compared with simple alternatives, not only with doing nothing. If an automated function saves five minutes but creates reviews, errors and complaints, the full balance may differ from the one described in advertising.

Sustained benefit is not the same as frequent use#

Downloads, open accounts, screen time and clicks may describe activity, but they do not prove benefit. In settings affecting minors, responsible review should connect use to an understandable aim: learning a skill, accessing information, reducing a barrier or enabling safe communication. That aim needs indicators that do not confuse quantity with improvement. More videos watched do not necessarily mean deeper understanding, and more alerts sent do not mean greater care.

People using a system can provide information that no dashboard summarises. Pupils may explain whether recommendations make sense; families may identify costs, timetable conflicts or barriers to exercising rights; staff may show where errors collect. Listening to these experiences does not turn an impression into evidence by itself, but it identifies what should be measured and which groups disappear from aggregate data. A sustainable improvement should persist without demanding constant surveillance or imposing a disproportionate burden on those who receive less support.

Costs belong in the same conversation as benefits#

Assessing only the visible function leads to incomplete decisions. There are financial costs, including licences, devices, training, support and replacements; environmental costs related to energy, equipment renewal and waste; and human costs such as interruption, anxiety, advertising exposure or time lost resolving incidents. There are also opportunity costs: resources spent on a platform that can no longer support personal assistance, libraries, mediation or activities that work better away from a screen.

Naming these costs does not mean that every innovation is a mistake. It means an institution should publish the balance it uses and explain who reviews it. If a tool depends on advertising, profiling or sensitive data, that fact should carry weight, particularly where minors are involved. If it requires families to buy equipment or pay for connectivity, equivalent alternatives need to be planned. Savings promised by automation must not hide the work of correcting wrong decisions, handling complaints and supporting people unable to use the service.

Evidence should be independent, understandable and situated#

A report prepared by a company selling a solution may contain useful data, but it is insufficient as the only evidence. It matters to know how participants were selected, how long outcomes were observed, what comparison was used, which limits the authors acknowledge and whether another body could review the methods. Success stories deserve attention, but a selection of favourable cases cannot be generalised to every school, neighbourhood, age or need.

Evidence must also be understandable to people living with the consequences. Responsible organisations can explain in plain language what is known, what remains uncertain and which decision is being made while data are missing. When a tool affects grades, benefit access, content moderation or health, the demand for explanation should be higher. A prudent principle is not to extend at scale a system whose effect on rights cannot yet be clearly described, and to maintain human review able to correct harm.

Maintenance, updates and an identifiable responsible person#

A technology does not end on its purchase or launch date. Terms change, vulnerabilities appear, algorithms are altered, devices age and users’ needs change. Before adoption or renewal, it is essential to know who will maintain settings, what support will cost, how an incident will be reported and how quickly it will be corrected. A contract or policy that assigns no responsibility leaves families and professionals without a clear route when something fails.

Planning should include initial and recurring training, accessibility, backups where appropriate and response procedures for failures. In a school, for example, showing a function is not enough; there must be guidance for an impersonated account, a platform that blocks someone by mistake or a person who needs a non-digital alternative. System quality is visible in its response on a difficult day, not only in a presentation. Maintenance should be budgeted and visible rather than dependent on one person’s goodwill.

Leaving a tool can also be a responsible decision#

Some solutions stop adding value, become too expensive, fail to respect privacy or simply no longer fit the need that justified them. Keeping them from habit can increase harm and make an orderly exit harder. Every adoption should therefore include exit criteria from the beginning: which signals will show that benefit has declined, who can suspend use, how affected people will be informed and how data will be recovered or deleted safely.

A responsible exit is not a defeat. It may follow learning that another option is more inclusive, safer or more useful. Transition needs time, support and honest communication. Families and young people should know what changes, what will happen to their data, where support is available and how an essential activity will continue. When an institution plans an exit, it avoids letting technical dependence make the community captive to a supplier or to a decision made during a fashion.

Periodic review protects without stopping learning#

Review does not need to be an impossible procedure. A clear schedule can include an initial check, listening to users after three or six months and an annual evaluation comparing benefits, incidents, costs, accessibility and rights. Conclusions should identify concrete changes: continue with adjustments, limit some uses, request supplier guarantees, provide alternatives or remove the tool. Publishing a summary lets trust rest on explainable decisions rather than reassuring phrases.

This approach gives young people and families a valuable lesson: technology can be used, questioned and changed. It is not an inevitable destiny or an object to celebrate merely because it is new. A mature community can recognise an advantage, require repair after an error and change direction when evidence advises it. Innovation gains legitimacy when it accepts that conversation and when responsible people show that wellbeing and rights weigh as much as novelty.

Sources for checking and further learning#

  • UNESCO, Recommendation on the Ethics of Artificial Intelligence and governance resources.
  • UNICEF, guidance on children’s rights in digital environments.
  • European Commission, information on the Digital Services Act and protection of minors.
  • Spanish Data Protection Agency, materials on minors, privacy and accountability.
  • OECD, principles on artificial intelligence and impact assessment.

These sources provide public frameworks for asking about risk, rights, oversight and repair. Consulting them does not replace analysis of each case, but helps a technology decision rest on visible and reviewable criteria.

Deciding through shared judgement#

A better decision does not require absolute certainty, but it does require a method that leaves a record. It is useful to note the purpose, data reviewed, open questions, date of the next review and the person responsible for hearing incidents. That record makes it possible to compare what was promised with what happened and prevents a tool from remaining simply because nobody looked again. When judgement is shared with families, young people and professionals, risks and needs appear that an initial demonstration had not shown. Continuing or withdrawing a technology then becomes an explainable, correctable decision directed towards the common good.

Questions that remain useful#

At every review, the same practical questions deserve a clear answer. Is the original need still present, are the claimed results still visible, and are people able to obtain support without unnecessary delay? Has the provider changed its terms, its data practices or the cost of the service? Are children, families and staff able to choose an accessible alternative? Recording answers makes comparison possible at the next review. It also prevents a temporary fashion from becoming a permanent rule merely because changing it seems inconvenient. Careful questions make responsible change possible.