Dr Claire Hudson continues a series of case studies derived from a BILT Associate Project: Staff Adoption of Generative AI (GenAI) in Teaching. These examples are suitable for staff with very limited GenAI experience.
Teaching context
Creating effective assessments is a challenging and time-consuming aspect of teaching. It requires alignment with learning outcomes, appropriate levels of challenge, and clear marking criteria, alongside consideration of institutional strategy including inclusivity, authenticity, and integrated assessment. With this in mind, assessment practice is evolving, with more emphasis on scaffolding and integration of formative assessment to support feedback literacy and ongoing student development.
Generating new assessment ideas stretches both time and creativity, therefore GenAI could help, not as a replacement for academic judgment, but as a tool to spark ideas, support iterative development and increase efficiency in assessment design. Drawing on my own practice and insights from staff interviews, this case study explores practical ways GenAI could be used to support assessment development
Top tip: When using GenAI for assessment, you should use your institution’s recommended platform, MS Copilot at the University of Bristol, for enhanced data protection.
1. Creating assessment questions from teaching materials
Teaching content can be uploaded to MS Copilot in PDF form, and used to generate assessment questions. Colleagues have experimented with this approach for essay-style questions, and while most don’t use the output directly, they report it is a fantastic way to generate multiple ideas when your “mind goes blank”. Overcoming this initial block is often enough to get your own creative juices flowing……
Similarly, GenAI can generate MCQs very quickly from your teaching material, allowing you to cherry-pick, review and refine the best ones.
Top tip: be specific with your prompt, you need to include pedagogic intent, not just topic coverage. Include:
- Intended Learning Outcomes (ILOs): to ensure appropriate depth and focus.
- Cognitive level: explicitly state required thinking (e.g. analyse, evaluate, synthesise) to prevent overly simplistic questions.
- Student level
- Assessment constraints: word count, use of literature, coursework or exam.
- Context/purpose: summative or formative, integration with other units/assessments, links to real-world context.
- Specify output: how many ideas, whether they should cover the same or different ILOs
Example prompt:
I am designing a summative coursework essay assessment for a [student level] unit in [topic]. Students have covered [insert topics or upload teaching material].
The intended learning outcomes are:
- [insert ILOs]
The essay should target higher-order thinking (e.g. analysis, evaluation, synthesis) and be 2000 words in length.
Generate 5 essay titles that:
- Encourage critical evaluation rather than description
- Require integration across several topics e.g. [examples]
- Have a clear relevance to [context anchor]
- Are sufficiently focused to be answered within the word limit
Have a go! You can even ask GenAI to quality control its own output, for example: “Please critique these titles against the ILOs and suggest improvements.”
2. Checking clarity of assessment wording
As educators, we are very familiar with our subject matter and ILOs. This familiarity can make it difficult to see ambiguity in our own wording; what feels clear to us may be interpreted in several ways by students reading the task for the first time.
Issues such as unclear command verbs, implicit expectations, overly complex phrasing, or assumed prior knowledge can all create confusion, leading students to misinterpret the task and perform poorly. Clarity is essential if we want to make assessments accessible and fair.
GenAI can be used as a“first-pass reader,” simulating how a student might interpret the question. GenAI could be asked to:
- identify ambiguous wording,
- explain what a student is being asked to do,
- check alignment between the question and the ILOs
3. Developing explanations and model answers
Several staff members interviewed had used GenAI to generate model answers or explanatory feedback for assessment questions. When collating a bank of objective task questions, several were submitted from colleagues without a rationale (explanation feedback). The interviewee stated: “I could have spent time working it out, but that’s not a good use of my time”. Instead, they asked GenAI to create these explanations, which were then verified by the original author.
Similarly, others described uploading lecture material alongside an assessment brief to create a skeleton model answer. One interviewee commented: “It provides a really good outline that you can quickly go through and edit a few bits and bobs. It’s 80% of the work done in a second. So that’s saving me a lot of time.”
In these examples, GenAI is used not to replace academic input, but to accelerate the production of drafts that are then refined and quality-controlled by academics.
4. Making marking criteria task-specific
This final application complements the conception and quality control described above. Students can find university-level marking criteria quite abstract, and therefore struggle to interpret what they look like in a specific assignment.
I try to contextualise the marking criteria so they are task- and discipline-specific, and GenAI can help with this. This ‘interpretation’ of the marking criteria can help students understand our expectations, particularly with non-traditional assessment types.
Top tip: It’s all in the detail! Provide both the marking criteria and the assessment instructions in your prompt.
Include:
- Assessment instructions (full details regarding format and expectations)
- Student level
- Marking criteria
- Output format
Example prompt
I am designing a [assessment type] for [student level] students in [topic].
The assessment instructions are [insert full instructions] and these are the marking criteria across grade bands [insert criteria]. For each criterion, write 1–2 concise sentences describing what a high-quality submission (70–100 level) looks like in practice in this assessment type.
Important instructions:
- Focus only on what this looks like in this type of assessment
- Do not refer to the specific assessment topic
- Do not repeat or paraphrase the generic marking criteria
- Translate expectations into clear, student-friendly language.
I don’t ‘translate’ the criteria for each grade band (e.g. Distinction, Merit, Pass, Fail), but you could definitely include this request, which could be helpful to markers as well as students.
Final thoughts
These examples describe how AI can support your own knowledge and creativity with assessment development. As one interviewee reflected: “not all the output is useful, but it does the heavy lifting!”
It is a good idea to seek institutional guidance before using GenAI for assessment, and most importantly: check and double-check all output for accuracy!
You may be thinking that some of the assessment types mentioned here, particularly essays and MCQs, are very vulnerable to student use of GenAI. You are right! However, MCQs still have clear formative value, and the essay example is really just a starting point. GenAI also provides the opportunity to rethink assessment; it can help mould your ILOs, teaching content and marking criteria into more creative and diverse assessment types. Use your imagination, or let GenAI provide it a spark!
Leave a Reply