Designing responsible AI feedback in higher education: faculty readiness, student judgement, and metacognitive calibration

Abstract

Artificial intelligence (AI) is now part of everyday academic work in higher education, including feedback. However, AI feedback does not automatically lead to learning. It can help students rethink a concept, improve a draft, and notice gaps in their work, but it can also create a false sense of progress when students accept suggestions without understanding them. This Perspective article argues that responsible AI feedback depends on faculty readiness and students’ evaluative judgement. Readiness is not mainly technical. It involves pedagogical judgement: knowing how to design tasks in which students work with feedback, compare it with evidence, revise with purpose, explain their choices, and remain responsible for the final submission. The article focuses on generative AI feedback and uses metacognitive calibration as a central concept. Calibration refers to the relationship between students’ confidence or self-judgement and independently evaluated evidence of performance. Evaluative judgement concerns students’ capacity to judge the quality and relevance of feedback and work, whereas calibration concerns the accuracy of their judgements relative to independently evaluated performance evidence. The article shows how ideas from feedback literacy, evaluative judgement, and self-regulated learning need to be adapted when feedback comes from systems whose accuracy and educational value cannot be assumed. It proposes a conceptual responsible AI feedback design cycle linked to assessment integrity, university policy, and faculty development. Although the cycle requires empirical testing, it offers a practical way to support reflection, calibrated judgement, and accountability without replacing students’ thinking.