Toward Sustainable Learning With AI : A Methodological Agenda for Accountable Coursework

As AI tools such as ChatGPT become a routine presence in student coursework, institutional responses continue to oscillate between pedagogical adoption and academic integrity enforcement. This polarisation leaves an important methodological gap: how to support legitimate AI use without allowing learning to collapse into unexamined output production. This paper advances a methodological agenda for sustainable learning with AI, centered on making AI‐mediated coursework accountable rather than merely detectable. We propose four methodological commitments for task design and evaluation: (i) treating AI use as a first‐class object of inquiry rather than a hidden variable; (ii) shifting evaluation from authorship policing to accountable reasoning and decision‐making; (iii) foregrounding boundaries, assumptions, constraints and trade‐offs as core learning targets; and (iv) designing review routines that privilege revision, justification and critique over polished final artefacts. To support adoption, we provide design heuristics and review prompts in tabular form, offering practical guidance for task design and evaluation that can be adapted across courses without reducing learning to simplistic metrics. The agenda reframes the central question from ‘Did students use AI?’ to ‘What forms of learning remain sustainable when AI is available and under what evaluative conditions?’