RESILIENT, EVIDENCE-BASED ASSESSMENT LAB
Real evidence.
Resilient assessments.
When AI can produce the answer, what shows that a student has learned?
REAL helps educators examine assessment vulnerabilities and turn the findings into practical changes to tasks, marking and assessment conditions.
ILLUSTRATIVE ASSESSMENT
Beyond the
finished answer.
A well-written recommendation
↳ What evidence supports the thinking?
Reward the learning that matters.
01 / THE QUESTION BEHIND REAL
What does the
assessment actually show?
An excellent submission is useful evidence only if it demonstrates the learning you intended to assess.
Generative AI can contribute to polished writing, analysis and recommendations. REAL examines where a student could use that assistance to bypass the intended learning, and where the assessment already provides stronger evidence of understanding.
We use AI resilience to mean how well an assessment continues to provide credible evidence of student learning when generative AI is available.
AI resilience depends on the task, how it is completed, what is assessed and how achievement is verified. No single feature makes an assessment AI-proof.
02 / THE REAL FRAMEWORK
Four principles.
One question about learning.
Reasoning, Evidence, Application and Learning provide a shared language for examining and strengthening assessment design.
R / REASONING
Make the thinking visible.
Require students to explain the decisions, interpretations and processes behind their response. This gives assessors more to judge than the finished product alone.
Can you see how the student reached the answer?
ILLUSTRATIVE REDESIGN
Recommend an intervention to improve employee wellbeing.
Recommend an intervention, then explain which two case facts shaped your decision and why you rejected one alternative.
An example to adapt to your learning outcomes. Stronger instructions work best alongside appropriate marking and verification.
Does this assessment provide real evidence of learning?
REAL03 / FROM AUDIT TO REDESIGN
A clear process.
Practical decisions.
Auditors conduct the review. Lecturers bring their disciplinary expertise and choose changes that fit their assessment.
- 01LECTURER + AUDITOR
Understand the assessment
Share the brief and any rubric or supporting materials. Establish the learning outcomes and the conditions in which students complete the task.
- 02AUDITOR
Review the evidence
Examine AI access, assessment design and opportunities for AI to substitute for the intended learning. Record findings against the assessment materials.
- 03LECTURER
Explore and decide
Understand each finding and its significance. Incorporate the recommended change, choose an alternative or adapt it to your context.
- 04LECTURER + AUDITOR
Redesign with purpose
Revise instructions and marking criteria using worked examples and tailored guidance. Bring agreed decisions together in a practical redesign plan.
04 / INSIDE THE REAL LAB
A workspace that helps
findings become action.
The REAL Lab is our assessment audit and redesign platform. It brings evidence, explanations, lecturer decisions and practical guidance into a structured journey.
See the finding.
Understand the reason.
Each finding connects to evidence in the assessment materials, a plain-language explanation and a practical recommendation.
This is an illustrative walkthrough of the platform experience. No assessment data is uploaded or analysed here.
The recommendation is assessed.
The reasoning is less visible.
“Present your final recommendation.”Illustrative brief · Task instructions
Students could submit a convincing conclusion without explaining the judgement behind it.
Assessment audit report
Findings, evidence and priority recommendations in one record.
Assessment redesign plan
Selected changes, their rationale and practical guidance for the task.
Rubric guidance
Suggested criteria and descriptors aligned with the revised assessment.
A FEW USEFUL ANSWERS
Before we begin.
Clear expectations are part of good assessment design. They are part of our approach, too.
Does AI resilience mean banning AI?
No. The appropriate role of AI depends on the learning outcomes and institutional policy. REAL focuses on whether students still demonstrate the knowledge, skills and judgement being assessed, wherever AI is permitted.
Does REAL detect whether a student used AI?
No. REAL reviews assessment design and conditions. It does not establish whether an individual student used AI or determine academic misconduct.
What if I do not have a marking rubric?
A rubric is helpful but optional. The audit can begin with the assessment brief, learning outcomes and available marking guidance. Any limits in the available evidence should be recorded.
Will the redesign create more marking?
The aim is proportionate change. A short explanation, a targeted verification step or a change to existing marking criteria may be more suitable than adding another submission. Lecturers can choose or modify recommendations to suit their context.
Which assessment formats does REAL support?
The current prototype supports written assessments, multiple-choice quizzes and presentations through separate assessment protocols.
Can a higher resilience result guarantee an AI-proof task?
No. Results support professional judgement about assessment design. They do not guarantee authorship or prove that AI cannot complete a task. Reasoning, contextual application and evidence can also be generated by AI, so completion conditions and verification remain relevant.
LET’S LOOK AT YOUR ASSESSMENTS
Start with a conversation.
Tell us about the courses you teach, the assessments you use and where you would value support.
Discuss an assessment review teixieradulalarthur@gmail.comReasoning.
Evidence.
Application.
Learning.