The feedback box: What is educational in the age of artificial intelligence?

Feedback is a neutral mechanism. The same response loop drives laboratory animal conditioning, commercial digital platforms engineered to capture attention, and classroom instruction. Because it is indifferent to its outcome, feedback does not automatically educate; it can just as easily condition or manipulate. Drawing on R.S. Peters, this paper argues that a feedback system becomes educational only when it meets a demanding set of conditions. It must engage the learner intentionally, with the learner conscious of what they are trying to master. Its methods must be morally defensible, ruling out coercion, manipulation, and indoctrination as means. It must aim at knowledge and understanding rather than only trained performance, giving learners the reason why things are as they are. What is learned must transform the learner's judgement and outlook rather than remaining inert. It must support the all-round development of the person rather than narrowing them to instrumental function and it must aim at autonomy, so that the scaffolding can eventually be withdrawn and the learner can think, judge, and act for themselves. These conditions are under direct pressure from contemporary AI systems, which can capture attention through addictive design and replace reasoned feedback with token economies that reward compliance rather than understanding. The school classroom is in a stronger position against these pressures than is usually recognised, because its boundaries are visible, its standards are publicly contestable, and its feedback is anchored in accountable human interaction that cannot easily be displaced by generative output. Higher education is far more exposed, precisely because of the freedoms that have always defined it: students working alone, constructing their own paths through reading and inquiry, and submitting artefacts that generative AI can now produce directly. The task is therefore not to adopt or reject AI but to configure feedback systems, in classrooms and university courses alike, so they remain answerable to educational purposes rather than to whatever the technology most easily affords.