The donkey and the algorithm: cultivating judgement infrastructure for AI literacy in higher education

Purpose This paper reframes artificial intelligence (AI) literacy in higher education as a problem of judgement formation rather than tool fluency alone. It asks how students can be supported not only to use generative AI, but also to evaluate, limit, disclose or refuse its use responsibly under conditions of uncertainty. Design/methodology/approach This paper develops a conceptual and pedagogical argument through engagement with AI literacy scholarship, evaluative judgement, information judgement and educational theory. Rhetorical literacy, informal logic and argumentation, and ethical reasoning/phronesis are used as a pragmatic framework for analysing the judgement capacities required for meaningful AI engagement. Findings This paper proposes a judgement-infrastructure account of AI literacy. It argues that student suspicion, selective use, uncritical reliance and refusal can be read as sites where judgement is practised, bypassed or left unarticulated. The Judgement-Infrastructure Gap Model is offered as a diagnostic heuristic for examining where judgement is supported or neglected across curricula, assessment, institutional policy and technological platforms. Research limitations/implications As a conceptual paper, the argument and model require empirical and design-based testing. Future research should examine how students justify AI use, limited use, disclosure and refusal across disciplines, and how programme-level scaffolding, assessment design, institutional policy and technological platforms support evaluative judgement. Practical implications AI literacy initiatives should include assessable activities that make reasoning visible, such as claim-evidence mapping, annotated AI outputs, oral defences, AI-free checkpoints, ethical decision logs and justified decisions about AI use, limitation, disclosure or refusal. These practices should be supported at programme and institutional levels rather than placed only on individual teachers or students. Originality/value This paper contributes a judgement-infrastructure account of AI literacy that treats both use and non-use as teachable and assessable forms of evaluative judgement. It extends existing AI literacy frameworks by linking student suspicion, refusal, assessment design, programme-level scaffolding and educational ecosystem conditions to responsible AI engagement.