The technological future of school is not decided when a new tool arrives. It is decided when a school defines what the tool is for, which learning it protects and which boundaries it will not cross. Families do not need to choose between enthusiasm and rejection. They need to know what their children will learn, which decisions remain human and how harm can be corrected. This guide offers a practical framework for placing technology, artificial intelligence and data at the service of education that is fairer, understandable and humane.
Picture an ordinary school meeting. Staff announce a platform that can adapt exercises, detect difficulties and save teachers time. A father asks what data it collects; a student wants to know whether she can make a mistake without being permanently labelled; a teacher worries that an automated recommendation will shape expectations. All three questions are educational, not merely technical. A purchase may take weeks, but its rules create habits for years. The first decision is therefore not which product to buy, but what kind of relationship with technology the school intends to teach.
Define the educational purpose first#
A tool deserves a place in the classroom when it improves an objective staff can explain without naming the brand: practising with timely feedback, making text accessible, collaborating remotely, representing a difficult phenomenon or freeing time for conversation. “Modernisation” is not a testable objective. Before a pilot, write down the existing problem, how it is addressed now, whom it affects and what evidence would demonstrate improvement. If the answer is only speed or novelty, the project is not mature enough.
A tool deserves a place in the classroom when it improves an objective staff can explain without naming the brand: practising with timely feedback, making text accessible, collaborating remotely, representing a difficult phenomenon or freeing time for conversation.
The school must also identify what cannot be delegated. An application can organise evidence, but it cannot understand a student's history by itself. It can suggest activities, but should not determine a major accommodation, sanction or high-impact grade without human review. This boundary should be written down and known to learners and families. Responsibility cannot disappear behind “the system said so”. There must always be a person with enough authority to explain, review and change a decision.
Purpose becomes practical when connected to a learning sequence. If artificial intelligence is used in writing, students need to formulate a question, test the response, find omissions, recognise uncertainty, cite support and reconstruct an answer in their own voice. Requesting polished output does not demonstrate competence. The valuable evidence is the process: drafts, choices, discarded sources, corrections and a short oral explanation. Used this way, the tool expands thought rather than replacing it.
A curriculum that teaches capability and limits#
Digital literacy is not simply the ability to operate applications. It includes understanding that an interface directs attention, a model can produce a persuasive falsehood, data enable inferences and a recommendation reflects objectives selected by somebody. A competent learner can use a system, but can also stop, choose an alternative and ask who benefits. Training solely for greater technological use turns dependence into the appearance of skill.
The curriculum can be organised around six progressive capacities: search and verification; creation and attribution; protection of data and identity; communication without harm; understanding automation; and preservation of autonomy. Primary pupils work with visible examples and simple choices. Secondary students add source comparison, bias, business models, authorship and collective consequences. A separate subject is not required for everything. Languages, science, tutorials, art and citizenship provide authentic contexts in which to practise.
Limits are also taught through design. A coherent school reserves screen-free periods when handwriting, handling materials, moving, observing or face-to-face discussion is more valuable. This does not turn paper into a moral symbol; it selects the right medium. Learners gain insight when teachers make that choice explicit: “We are not using devices today because we need to compare ideas in person,” or “We are using one because this simulation shows something our laboratory cannot.”
Understandable privacy and minimum data#
Data protection begins before a consent form. The school must inventory what information enters each service, for what purpose, for how long, where it is stored, who can access it and how it is deleted. Names, voices, images, writing, browsing behaviour and learning patterns may reveal more than expected. Data minimisation supplies a useful rule: if a feature works without identifying a child, identity should not be required; if work does not need to be retained, it should be deleted.
Families deserve a short explanation in everyday language. They need to know whether a service is compulsory, which alternative exists, whether student content is used to train models and how rights can be exercised. Consent given when participation is otherwise impossible does not solve the problem. The Spanish Data Protection Agency's guide for educational centres clarifies responsibilities and common processing involving images and communications. Schools should turn legal duties into procedures a family can actually use.
With generative services, one precaution is easy to remember: do not enter full names, histories, reports, family conflicts or identifiable work into unapproved services. Staff can experiment with fictional cases or anonymised data. Teachers need a channel for asking before they improvise, rather than after an incident. A short list of permitted, uncertain and prohibited uses often protects people better than a long policy no one can recall.
Assess without turning prediction into sentence#
Automation can identify patterns, but a pattern is not an explanation. A system predicting dropout risk might prompt timely support; it might also lower expectations and increase surveillance of students who already face disadvantage. A prediction must therefore never become a pathway, sanction or exclusion on its own. Staff need to examine its variables, false positives and whether errors are concentrated among disabled students, speakers of another language or learners with fewer resources.
In assessment, closed-answer practice with low consequences can more readily use automated marking. An essay, creative work, suspected cheating or a final decision requires human judgement and an opportunity to respond. Students should know which part involved a tool and what evidence supported the outcome. If challenged, review must be accessible, prompt and conducted by someone able to change the decision. An appeal route without authority to rectify anything is just a mailbox.
Assessment must also adapt to the presence of AI. A general ban creates rules that are hard to apply; unconditional permission hides who did the thinking. Define use for each task: prohibited, allowed for brainstorming, allowed with a process record, or integrated as an object of criticism. Authorship can be examined through conversations, versions and justified decisions. This teaches intellectual honesty without making invasive monitoring the automatic response.
Equity, accessibility and the right to an alternative#
An innovation is not inclusive merely because it is online. Some homes share one device, have unreliable connections, conflicting schedules or less adult support. Some students require screen readers, captions, keyboard control, more time or lower sensory load. Before adopting a tool widely, the school should test it with those who encounter the greatest barriers. If families must buy equipment, print materials or provide technical support, the cost has been transferred rather than removed.
The right to an alternative must be equivalent, not punitive. A learner who cannot or should not use an application must reach the same objective without receiving a longer or less valuable task. Accessibility is tested with real people and situations, not just a supplier's tick box. The European Commission's DigCompEdu framework connects educators' digital competence with inclusion, assessment and self-regulated learning, reminding schools that pedagogical judgement is part of technological competence.
Equity requires disaggregated results. A positive average can conceal that one group takes twice as long, leaves more often or receives less ambitious recommendations. During a pilot, record access, requests for help, errors, appeals and perceived experience without collecting unnecessary personal information. When a disparity appears, pause and investigate. The urge to scale cannot outweigh the right to learn under fair conditions.
Govern a pilot before purchasing#
A responsible pilot can last four to eight weeks, cover one objective and name an accountable person. Before it begins, record the baseline: time spent, quality of learning, incidents and student experience. Decide which indicators justify continuation and which trigger a stop, including an access gap, inexplicable outcomes, excessive data collection or increased invisible work. Setting these rules in advance reduces the pressure to defend an investment simply because it has begun.
The pilot should include a failure scenario. What happens if the connection disappears, the provider changes terms, a response is offensive or an account is lost? There should be educational continuity, an incident channel and a resolution period. Calculate total cost as well: licences, configuration, training, review, support, accessibility and exit. A cheap purchase may be expensive to leave when materials are locked away or staff become dependent on a proprietary format.
UNESCO's guidance on generative artificial intelligence in education recommends a human-centred, age-appropriate approach attentive to privacy. It does not offer a magical product list. It connects technical choices to rights and educational purpose. That provides a useful discipline for an evaluation group: ask what learning gains, which new risk appears and who genuinely has the power to correct it.
The practical role of families and students#
Families can request answers to seven questions: which educational objective the tool serves; what data it uses; whether it is compulsory; which alternative is available; who reviews its results; how a challenge is made; and when continuation will be assessed. They do not need to debate code or algorithms. They need intelligible rules. If a school cannot answer, the solution is not governed well enough. Discussion improves when it starts with a specific case instead of general positions for or against technology.
Students should participate before, during and after a pilot. They can detect unclear instructions, barriers and unforeseen uses adults miss. Participation does not give them final responsibility; it recognises their experience. A small, diverse group can test tasks, propose language for rules and explain what helps or distracts. Afterwards, the school should report which suggestions it adopted and why others were declined. Asking for opinions without responding undermines trust.
At home, a useful conversation avoids interrogation. Questions such as “Which part did you do?”, “Which response did you verify?”, “When did you decide not to use it?” and “Who would help if it made a mistake?” reveal the process. In a dispute, keep the task, relevant screenshots and communications, and ask for an explanation of the criteria before alleging bad faith. The purpose is to restore learning and responsibility, not to win an argument.
A school agreement that can be revised#
Every school can summarise its policy on one visible page: objectives, accepted uses, decisions requiring a person, prohibited data, alternatives, review channel, accessibility and evaluation schedule. It should be updated through real examples and training. The European Commission's Digital Education Action Plan highlights both an effective digital education ecosystem and the development of competence. Both matter. Equipment without governance leaves learners exposed, while teaching without reliable access deepens inequality.
The final question is not whether schools will have technology; they already do. It is whether they will teach young people to live with it through autonomy, care and evidence. A sound technological decision can be explained, tested, challenged and reversed. It distributes benefits without hiding costs, protects people with less power to choose and preserves time for what only an educational community can provide: attention, relationships, judgement and hope. That is the future being decided today, one everyday rule at a time.
