A school can roll out tablets, an assignment platform or an artificial-intelligence assistant and still fail to innovate. The difference becomes visible when it tries to change supplier, teach through a service outage or recover students’ portfolios. If teaching stops, materials cannot be exported, or staff can no longer run the activity without that application, the tool has stopped supporting the educational project: the school depends on it. This chapter gives leadership teams, teachers and families a practical method for distinguishing genuine pedagogical improvement from technological dependence, both before a contract is signed and throughout the life of the service.
The scene is easy to recognise. For two academic years, a school uses a platform to submit work, comment on drafts and store assessment evidence. Teachers create hundreds of rubrics; students keep their projects there; families monitor progress. At renewal, the price rises and several functions move to a more expensive tier. School leaders ask how the information can be transferred elsewhere. The supplier can download grades to a spreadsheet, but not comments, versions of assignments or course structure. Legally, the school may own the data; operationally, it is trapped. The exit cost is no longer merely financial. It includes teachers’ time, institutional memory and continuity of learning.
Innovation improves a practice; it does not merely introduce a tool#
An educational innovation should be describable without naming a brand. For example: “students should receive useful feedback before submitting a final version”, “a pupil absent through illness must be able to continue learning”, or “teachers need to identify a misunderstood concept and adjust their explanation”. Such statements allow technological and non-technological options to be compared. If the objective can only be expressed as “deploy platform X”, the decision has started with the product. The simplest test asks which observable behaviour will improve, for whom, by when and according to what evidence. An attractive novelty may save clicks while doing nothing for understanding, autonomy, inclusion or the educational relationship.
Dependence does not mean using a tool frequently or paying for a licence. Textbooks and electricity are also necessary infrastructure. The problem begins when a school loses meaningful capacity to choose. Dependence is technical when files cannot migrate in usable formats; pedagogical when the application’s features dictate the learning sequence; organisational when only one person understands the configuration; contractual when leaving brings disproportionate penalties or work; and cognitive when teachers or students stop practising abilities they need beyond the system. Separating these dimensions prevents two opposite mistakes: rejecting all technology through fear, and accepting every novelty because it performs well in a demonstration.
UNESCO’s Global Education Monitoring Report on technology asks whether a solution is appropriate, equitable, scalable and sustainable, while insisting that technology should serve people and educational outcomes. That is more demanding than counting devices or logins. The European Commission’s Digital Education Action Plan likewise treats organisational digital capacity and community competence as conditions of effective digital education. Together, these references move the conversation away from enthusiasm about features and towards governance: who defines success, whether staff are prepared, how excluded students are protected and whether the institution can change course when evidence does not support the original promise.
Seven tests a tool should pass before entering the classroom#
First: purpose and comparison. The school should document the starting problem and at least one alternative. A baseline is useful: how long students wait for feedback, how many assignments are inaccessible, which mistakes recur, or how teachers describe their workload. A limited pilot can then be designed with comparable groups and an end date. Asking whether users “liked it” is not enough. Review learning, participation and total time, including configuration and support, as well as different effects on pupils with disabilities, poor connectivity, limited language proficiency or shared devices. A benefit that appears only because invisible labour is not counted is not sustainable improvement.
Second: verifiable export. Before purchasing, the school should request an actual test export rather than accept a sales promise. A downloadable file is not necessarily portable. A PDF may preserve appearance but make rubrics impossible to reuse; a table may lose relationships between tasks, comments and authors; a proprietary format may be unreadable without an active licence. A proper test imports a sample into another tool or opens it with common software, checking names, dates, permissions, versions, attachments and metadata. The contract should state who performs extraction, how long it takes, what it costs, what happens at termination and when the supplier deletes remaining copies.
Third: pedagogical continuity. Every critical function needs a reasonable alternative. If the platform is unavailable for a week, can students still obtain instructions, submit work and receive feedback? The fallback need not reproduce every convenience, but it must preserve the right to learn and to be assessed fairly. It may include regular local copies, agreed templates, an approved communication channel and rules for extending deadlines. It should be rehearsed like any other contingency procedure: choose one subject, deliberately disconnect the service for several hours and record what is missing. A plan that works only when the resident expert is available is not yet an institutional plan.
Fourth: pedagogical control. Teachers must be able to disable recommendations, alter sequences, correct outputs and explain decisions. With adaptive or AI systems, it matters which data informs a recommendation and what happens when it is wrong. No automated score should independently become a grade, referral or label about a child’s ability. Human oversight is not the act of clicking “approve” at the end. It requires time, understandable information and authority to contradict the system. Schools should also retain some activities without automation so that writing, calculating, remembering, discussing and revising remain students’ capabilities rather than functions silently outsourced to a provider.
Fifth: access and inclusion. A tool considered innovative for most users may create harmful dependence for someone who uses a screen reader, needs captions, shares a phone or relies on mobile data. Those students should participate in the pilot from the beginning. Tests should cover keyboard navigation, contrast, text-to-speech, file size, unreliable connections and equivalent alternatives. “They can ask for help” is not equivalence if a pupil must repeatedly disclose a difficulty or wait for an adult. The right solution reduces barriers without creating a second-class route. Families also need an honest statement of required equipment and connectivity, including accessories, updates and storage space.
Sixth: data and accounts. The school needs a simple inventory: what information is collected, for which purpose, for how long, who can access it and which subcontractors are involved. It should avoid gathering data “just in case”, using teachers’ personal accounts or requesting unnecessary permissions from pupils. Leaders must decide how users are created and deleted, what happens when someone leaves, and how incidents are handled. Privacy is not solved by a consent box, especially in education where the supposed alternative may not be freely available. The operational question is whether the learning can be achieved with less data and whether a dignified route exists when a family raises a well-founded concern.
Seventh: shared knowledge. If one teacher alone administers the system, apparent efficiency conceals fragility. The school needs short instructions, named deputies, training based on real situations and recognised maintenance time. Training cannot stop at buttons. It should cover when the tool is appropriate, how to identify a doubtful output and how to work without it. Each term, staff can identify activities that have become impossible outside the platform. If the list grows, the school should recover open formats, independent routines and transferable knowledge. Mature digital competence includes the ability to leave a technology, not only the ability to operate it.
A contract that preserves the ability to say “enough”#
Exit conditions must be negotiated before entry. Procurement documents should specify export formats, migration assistance, deadlines, maximum costs, service availability, notification of changes, retention and deletion, and the treatment of material created by students and teachers. They should prevent a unilateral update from suddenly changing the educational or commercial use of data. Wherever possible, schools should favour interoperable standards, institutionally managed authentication and licences that do not bind a learning history to a single account. Renewal should never happen through inertia. It requires a pedagogical and technical review based on evidence gathered from the school’s own experience.
A small committee can govern that review without creating heavy bureaucracy. Leaders contribute authority and resources; teachers describe practice; digital coordinators check integration and security; students expose invisible friction; families clarify access conditions and expectations; and, where relevant, data-protection or inclusion specialists assess specific risks. The supplier participates to answer questions, not to define success. A meeting should end with recorded decisions: continue, modify, limit or withdraw; who will act; and when results will be checked. Five questions can structure the record: what improved, who was excluded, which new cost appeared, can we leave, and what would we learn if the tool disappeared tomorrow?
Measuring the balance after the initial excitement#
At thirty, ninety and one hundred and eighty days, the pilot should be compared with its baseline. Measures should concern the educational purpose rather than vanity metrics such as open sessions. Record teacher time, support requests, outages, accessibility problems, complaints, duplicated tasks and the ability to recover materials. Qualitative evidence matters too. A short interview may reveal that students submit more work because reminders are persistent but revise less thoughtfully, or that one feature removes routine correction and creates time for better conversations. The purpose of evaluation is not to prove that the purchase was correct. It is to preserve institutional permission to recognise a mixed outcome and change it.
Families can use a domestic version of the same test. Instead of asking only how much screen time an assignment requires, they can ask what the child learns, what alternative exists, which data is supplied and whether the work can be kept. If an incident recurs, describe it factually —device, date, activity and consequence— and ask for the school’s procedure rather than looking for someone to blame. Students can participate as well. Learning to export a project, acknowledge AI assistance, protect credentials and continue a draft offline is part of digital autonomy. The best educational tool does not make the supplier indispensable; it makes the learner more capable.
An exit checklist before renewal is approved#
Before renewing, the school should be able to answer yes to six questions: it holds a usable copy of essential work; another person can administer the service; a fallback exists for critical functions; teachers can override automated decisions; families know the support routes; and a partial migration has been tested. A no should receive an owner and deadline. If several answers remain negative, limiting or removing the tool may be more innovative than keeping it. Leaving a solution that no longer serves its purpose is not failure. It demonstrates responsible governance and prevents sunk costs from determining the next educational decision.
The practical conclusion is simple, although it requires discipline. Innovation is a demonstrable improvement that expands the ability to teach and learn. Dependence is the loss of alternatives, knowledge or control until choice is no longer viable. The boundary lies not in the technology but in its conditions of use. A school can use a platform extensively and remain autonomous if it retains materials, competence, pedagogical judgement and a rehearsed exit. Conversely, it can depend on a rarely used application if something irreplaceable is stored there. Asking about exit before entry protects learning and allows the community to innovate with greater confidence.
Sources for checking and further reading#
The international framework is available in the UNESCO Global Education Monitoring Report on technology in education. For European strategy and institutional capacity, see the European Commission Digital Education Action Plan. These are starting points for setting criteria; each decision must also reflect applicable law, the school’s data-protection assessment and tests carried out with its own community.
