How a UK Master’s Programme Is Teaching Responsible AI Use

Getting students to “use artificial intelligence responsibly” is easy to say but hard to enforce. For international master’s students, many on intensive one-year programmes, the challenge of interpreting academic English and unfamiliar assessment conventions often heightens the temptation to hand too much of the work to a generative AI tool. Two academics from the University of Chester, Andrew Firr and Alex Fenton, have turned this dilemma into a teaching opportunity through the way they design their Level 7 engineering management module’s capstone assignment.

Students are set a practical design problem: they must examine a real campus process, identify where it slows down, gather observations and propose a feasible redesign in a 3,000-word essay. AI is permitted for planning and improving clarity, but it may not be used to fabricate the evidence that underpins the analysis. That means no invented queue measurements, no AI-generated screenshots, no unchecked citations and no diagrams of processes the student hasn’t actually observed. The assessment itself reinforces this boundary through a case study, a diagram, a fact-checking exercise and a mandatory short statement describing how AI was used.

The brief deliberately builds assessment literacy. Students paste part of the assignment into a generative AI tool to get a plain English explanation, then compare it with the official brief and marking criteria, learning to spot what the AI clarified and what still needs checking. By making the rules concrete rather than generic, the assignment removes guesswork and reduces the false confidence that polished AI text can create. Students still have to justify every claim and demonstrate where their evidence came from.

Why Chester’s Assessment Design Is a Game Changer for AI Policy

The explicit boundary between acceptable AI support and misuse

The module sets out precisely what students can and cannot ask AI to do. AI may explain where a process slows down, suggest questions for analysis or polish a student-written paragraph. It may not invent observations, create survey data, generate a fake screenshot, draw a diagram of something the student didn’t examine or supply unchecked references. If a student counts people waiting in a queue, that number is evidence that must be their own. This explicit list replaces vague warnings and gives students a clear framework for deciding where AI can assist and where their own work must stand alone.

Advertisement

Reconciliation checks as a safeguard against fabricated data

A distinctive feature of the Chester approach is the reconciliation check — a basic consistency test between different pieces of evidence. For instance, a student might compare an observed queue length with a waiting-time estimate to see if the two are plausible. Because AI cannot legitimately provide the real observations, any student tempted to fabricate data will find it difficult to pass this cross-check. The requirement turns the assessment into a task where genuine field evidence carries the weight, pushing students towards higher-order thinking rather than assembly of AI-generated text.

Why international students benefit most from this clarity

Many international master’s students arrive with strong subject knowledge but face a steep learning curve in UK academic conventions and the English-language subtlety of assessment briefs. The Chester module uses AI to bridge that gap — asking the tool to explain the brief in plain English, helping students grasp what “critical analysis” means in practice, and improving language expression in a paragraph the student has already written. By tying AI support to assessment literacy, the approach turns a potential risk into an equity lever without ever letting AI supply the core intellectual work.

What this means for the ongoing assessment debate

Across higher education, some advocate abandoning written assessments in favour of oral presentations or exams to guard against AI misuse. Firr and Fenton argue that a wholesale shift risks disadvantaging students who speak English as an additional language and ignores that many real-world tasks still require written outputs, often produced with AI assistance. Their model suggests retention and redesign of written work, with the emphasis shifting to process evidence, judgement and the ability to verify claims. Requiring an AI use statement — where students name the tool, describe how they used it, how they checked the output and which parts are their own — adds a reflective layer that encourages genuine academic integrity.

Adopting the Chester Model: A Practical Guide for Educators

  • Set explicit, granular boundaries. List exactly what AI may (e.g. explain the brief, suggest analytical questions, improve clarity) and may not do (e.g. invent observations, create graphics, supply unchecked references). Students need concrete examples, not general warnings.
  • Design tasks around real evidence. Build assessments that demand field observations, measurements or original data that AI cannot fabricate. The Chester module’s queue-count requirement illustrates how to make generative AI irrelevant for the core evidence.
  • Incorporate reconciliation checks. Require students to cross-verify different pieces of evidence — such as checking whether an observed queue length is plausible against a time estimate. This simple consistency test deters fabrication and reinforces analytical habits.
  • Mandate an AI use statement. Ask students to report which tool they used, how they used it, how they verified its output and which parts of the work depended on their own evidence and judgement. This acts as a reflective prompt and a gentle deterrent against misuse.
  • Use AI to build assessment literacy. Have students paste the assignment brief into a generative AI tool to get a plain English explanation, then compare it with the official brief and marking criteria. This trains them to recognise where AI adds value and where official documents retain authority — especially helpful for multilingual learners.