The list reached Nuria’s desk early in the morning, coloured as if it offered certainty: green, amber, red. She was a school counsellor, and the software had placed Ivan in red because of absences, missed assignments, and a recent drop in grades. The automated instruction was familiar: call the family, begin intensive monitoring, and restrict some activities until he was back on track. Nuria did not open the form first. Before break she saw Ivan by the library, his phone face down. When she asked how the week was going, he took a while to answer. His grandmother had just left hospital and his mother worked nights; for ten days he had been making dinner, helping with medication, and making sure his younger sister got home. None of that existed on the screen.
This does not mean school alerts have no use. An attendance signal can open support in time, and a sustained fall in performance deserves attention. The problem begins when a recommendation is handled as a portrait rather than an invitation to find out more. Human mediation is not the final gesture of pressing “confirm” after reading a prediction. It is the work of hearing reasons, holding disagreement, and answering for an exception when a category reduces a life to columns that cannot hold it.
A signal should prompt a question, not deliver a verdict#
Risk systems combine indicators because schools need to notice circumstances that can otherwise be missed. Absences, lateness, changes in performance, or silence on a learning platform may point to genuine difficulty. Yet every indicator has many possible causes. An absence may mean disengagement, a medical appointment, broken transport, family care, or an attempt to stay safe during a conflict. When a system keeps only the mark and drops the circumstances, it looks most objective precisely where it needs the most humility.
Risk systems combine indicators because schools need to notice circumstances that can otherwise be missed.
The practical difference lies in the verb attached to the alert. “This student is at risk” fixes an identity and prepares punitive responses. “This combination of signals needs a conversation” preserves uncertainty and requires someone to seek information the data did not contain. The second statement is not softer; it is more accurate. It allows a school to respond without mistaking a statistical likelihood for a proven reason. Whoever receives that hypothesis has a duty to test it and needs time, a route for listening, and authority to pause action.
What Nuria learns by listening#
Nuria did not need Ivan to disclose every private detail about his family to see that the first reading was incomplete. She asked what he could share, what he needed from the school, and what he did not want circulated among staff. Ivan said he did not want to become “the problem boy”; he needed a different deadline for two assignments, access to the library at lunchtime to call his grandmother, and no meeting that would force his mother to miss work. Those conditions did not erase the unfinished work, but they changed what a fair response looked like.
Listening does not mean accepting any account without questions. It means building a process in which a person’s explanation counts as relevant evidence, can be checked with care, and receives an answer. Sometimes motives will conflict: a family may want confidentiality while a team needs to coordinate support; a student may reject help the school considers important. Mediation exists to work through that disagreement, explain limits, and find a proportionate path, not to pretend a formula has already decided for everyone.
In this case, the counsellor and Ivan discussed what information he authorised her to share with his tutor. They separated the necessary fact — there was a temporary caring responsibility — from medical detail that had no reason to enter a record. The tutor adjusted deadlines, and the school checked whether a lunch grant and a family-support contact could reduce part of the burden. The alert became useful because it stopped operating as an order.
The counter-evidence sheet#
To ensure that a conversation does not depend on one person’s exceptional goodwill, a school can require a counter-evidence sheet before any action affects an opportunity: access to an activity, a course pathway, a sanction, a grant, or a referral. It is not paperwork designed to create another archive. It is a simple brake on the impulse to act with incomplete data.
The sheet should begin by recording the recommendation as it arrived: which system made it, which signals it used, when it was generated, and what action it proposes. It should then request information that may contradict, qualify, or explain those signals. Not merely “staff comments,” but direct questions: what circumstances does the system fail to measure? what has the affected person said? which sources are worth checking? what information would be unfair to demand? That last question is essential. Counter-evidence cannot become an interrogation that penalises someone for declining to expose private life.
A third section should set out a provisional response. If harm is possible while the case is being explored, what action reduces that harm without closing doors? Instead of refusing a place because of recent absences, the school might reserve it for a week and arrange a conversation. Instead of immediately referring someone to a disciplinary route, it might begin with reversible academic support. The document should name the responsible person, the review deadline, and the reason that would justify changing or keeping the decision.
Finally, the sheet needs room for the student’s disagreement. Ivan might write, dictate, or confirm a short sentence: “My absences relate to family care; I agree that my tutor can know that reason, but not medical details.” If he does not want to give an explanation, that cannot count against him. Respectful mediation recognises that cooperating with a process is not the price of fair treatment.
A conversation before intervention, with power to change it#
A required conversation protects only when it happens before a decision becomes irreversible and when it can alter the outcome. Contacting a family after an opportunity has been removed may gather information, but it is not mediation. The rule can be plain: where an alert will affect a meaningful educational option, explain the recommendation in ordinary language, hear the person, and record counter-evidence before deciding. Emergency exceptions should be rare, reasoned, and reviewed afterwards.
The conversation must fit the setting. Some students need a trusted adult; others need an interpreter, an appointment outside ordinary hours, or a written way to take part. A family working shifts cannot be excluded because the only meeting is offered at eleven in the morning. Accessibility is not a friendly extra. It determines who can correct a mistaken picture of themselves.
The person moderating also needs defined authority. If the counsellor finds a strong explanation but cannot change the classification or request that an intervention be paused, the system still has the final word. There should be a short route to review the recommendation, document the exception, and see whether the rule is producing repeated errors. Accountability begins with knowing who can say, “not yet.”
Learning from exceptions without building a suspicion file#
Counter-evidence sheets can reveal patterns a school should examine carefully. If many red alerts are explained by care duties, unstable housing, or family work schedules, resolving individual files is not enough. The school should ask whether the model treats precarity as lack of commitment, whether its thresholds punish people with less room to manoeuvre, and what institutional support is missing. The right lesson may be to change the system rather than demand increasingly intimate accounts from students already carrying difficulty.
That learning needs retention limits. Information gathered to correct an intervention should not remain attached forever to a minor’s record. It can be useful to keep the type of incident and the design lesson, while deleting sensitive detail once the support is resolved. Otherwise, a tool created for listening starts to build a memory of vulnerability that may cause harm later.
Each term, a team can review how many recommendations changed after listening, how long responses took, which actions were paused, and whether some groups receive more corrected alerts. Numbers do not replace stories, but they show whether mediation is real or decorative. A high correction rate does not necessarily prove staff failure; it may show that the system was making too many claims with too little context.
Nuria closed Ivan’s case without erasing the difficulty or turning it into a label. They agreed a date to review workload and deadlines; the alert remained a tested signal, not a diagnosis. The decisive act was modest: before intervening in his path, someone asked what was happening and allowed the answer to have consequences. That is the human task no screen can reduce to a click. A school’s next step is to pilot a counter-evidence sheet for high-impact alerts, guarantee a prior conversation, and examine which decisions changed because of what the model could not know.
