Metro, Tuesday 4 August 2026
A state school in Salamanca has introduced a humanoid robot, roughly a metre tall, to lead English lessons for pupils aged four to twelve, in what is reported as the first full-time deployment of its kind in a European classroom. The robot, developed by a Spanish edtech company at a reported cost of around £20,000, can hold basic conversation, correct pronunciation, and adapt its pace to individual pupils. The school's head teacher said the robot was intended to supplement rather than replace staff, and that early results on engagement had been positive. Some parents object, saying young children should be taught language and social skills by a human adult, not a machine, and that the school moved to install the robot without a meaningful consultation process. New York State has gone the opposite way. It passed a law this year prohibiting AI systems from serving as classroom teachers, following advocacy from unions and child development specialists who argued that early years education depends on relationships a machine cannot form.
Two education systems looked at the same technology and reached opposite conclusions about whether children should be taught by it. That divergence is not really about the robot. It is about who gets to decide, and on what evidence, when a cost saving and a stated pedagogical concern point in different directions.
New York's ban did not happen because a machine failed a trial. It happened because legislators were persuaded, in advance, that relationship-building in early education matters enough to rule the option out before it could be adopted piecemeal, school by school, under budget pressure that makes a one-off £20,000 purchase look attractive next to a teacher's salary. Salamanca had no such rule, so the decision was made locally, by a head teacher weighing engagement figures against a consultation process parents say never happened properly.
The manuscript's account of what schools actually transmit to children, distinct from what they claim to transmit, is relevant here in a way that goes beyond robots:
Education does not simply transmit knowledge. It transmits expectations about legitimacy. When rules are presented as universal but experienced as conditional, people learn that moral language is decorative and that power determines outcomes.
A child in the Salamanca classroom is not just learning English from a machine. They are learning, whether anyone intends to teach it or not, that a decision framed as being about them can be made without them, or the people responsible for them, having a meaningful say in it. That is the quiet curriculum described here: not the subject on the timetable, but the lesson embedded in how the subject got there. Parents who say they were not properly consulted are not objecting to English lessons. They are noticing, correctly, that the process which delivered the robot to their children's classroom ran without them.
Rules are introduced as fixed, then experienced as flexible. Expectations are stated clearly, then relaxed quietly. People are told what should happen while learning what will happen.
Consultation is exactly this kind of rule. It is stated as a requirement, understood by parents as a fixed part of how school decisions get made, and then, in this instance, experienced as something that could be compressed or skipped when a decision had already effectively been taken. The parents were told what should happen. What they learned was what would happen.
None of this requires the robot to be a bad teaching tool. Early engagement figures may well be genuine. What the New York comparison shows is that "does it work" and "should it be adopted this way" are different questions, and only one jurisdiction here has actually attempted to answer the second one in advance, on behalf of children too young to be consulted about either.
This entry sets a jurisdiction that has legislated against the practice against a school that adopted it regardless, and asks what the children in each system are actually being taught about who decides.