Assessment Design Strategies
Effective assessment design in the age of GenAI should support the development of AI literacy while ensuring that students can still demonstrate independent knowledge, disciplinary understanding, and professional competence. Assessment tasks should therefore be structured to make student thinking, decision-making, and learning development visible.
Strategies that support this include:
1 Multi-stage Tasks
Students submit plans, drafts, checkpoints, or progress updates before a final submission, demonstrating the evolution of their thinking.
2 Oral Components
Presentations, vivas, demonstrations, or Q&A discussions, enabling students to explain, justify, and critically reflect on their work.
3 Reflective Elements
Students articulate their reasoning, learning processes, and — where relevant — how they used, evaluated, or chose not to use AI tools.
4 Collaborative Work
Interaction, negotiation, and shared problem-solving, supported by mechanisms for recognising individual contributions.
5 Authentic, Real-World Problems
Contextual judgement, application of disciplinary knowledge, and the ability to navigate complexity rather than reproduce information.
These approaches shift assessment away from reliance on a single final artefact and towards evidence of the learning process.
Assessment Balance
A coherent assessment strategy should maintain a deliberate balance between:
Track 1 — AI-Restricted Tasks
Students must demonstrate foundational knowledge, core concepts, or professional competencies independently, without AI mediation.
Track 2 — AI-Integrated Tasks
Critical evaluation of outputs, ethical and transparent use, and awareness of bias and limitations.
This balance ensures that assessment both safeguards essential disciplinary learning and supports the development of future-ready, critically AI-literate graduates.
Assessment Strategies for Large Cohorts
Designing valid and meaningful assessment for large student cohorts presents particular challenges, including workload, scalability, and the frequent requirement for anonymous marking. In AI-enhanced contexts, these challenges are intensified by reduced visibility of individual learning processes.
However, effective assessment redesign at scale does not require intensive individual interaction for every student. Instead, it relies on structured design choices that make learning visible while remaining feasible.
Staged and Structured Assessments
Staged assessment design allows learning to be evidenced over time without substantially increasing marking load. Early stages may be formative or completion-based rather than fully graded.
Examples:
- Proposal or plan submission (pass/fail or checklist-based)
- Annotated outline or draft with targeted feedback prompts
- Final submission assessed against full criteria
Checkpoints and Low-Weight Touchpoints
Short checkpoints embedded within the assessment timeline can improve transparency and engagement at scale.
Examples:
- Brief reflective prompts on decision-making or challenges encountered
- Short self-check questions linked to assessment criteria
- AI-use disclosure or reflection statements (where relevant)
Such checkpoints can often be reviewed quickly or sampled rather than fully graded, helping to maintain feasibility.
Peer and Self-Evaluation
Peer and self-evaluation approaches can be particularly effective in large cohorts when clearly structured and supported.
These may include:
- Peer feedback templates focused on criteria rather than grades
- Self-evaluation checklists submitted alongside final work
- Individual contribution statements in group assessments
When aligned with marking criteria, these approaches encourage responsibility for learning and provide additional evidence of engagement without adding to staff marking workload.
Designing for Scalability
Across large cohorts, small structural changes can have significant impact.
Emphasising process evidence, structured staging, and learner reflection enables assessment that is: more resistant to inappropriate AI use; more transparent and equitable; more sustainable for staff.
In large cohorts, clarity, structure, and intentional design are more effective than increased surveillance or individualised scrutiny.
Large Cohort Strategies
| Design Challenge | Assessment Strategy | Purpose / Benefit |
|---|---|---|
| Limited visibility of learning processes | Staged assessments (proposal, draft, final submission) | Makes learning development visible over time; reduces reliance on a single artefact |
| High marking workload | Completion-based early stages (checklists / pass–fail) | Supports process without significantly increasing marking time |
| Large numbers prevent individual interaction | Low-weight checkpoints (short reflections, decision prompts) | Provides insight into thinking and engagement at scale |
| Risk of generic or AI-generated submissions | Targeted reflective prompts (reasoning, challenges, AI use) | Encourages metacognition and personalised responses |
| Difficulty evidencing individual contribution | Self-evaluation statements | Supports accountability and reflection without additional grading |
| Group work in large cohorts | Structured peer evaluation templates | Surfaces contribution and collaboration skills |
| Scalability concerns | Sampling approaches (selective review or oral clarification) | Maintains integrity signals without full cohort interaction |