How Dutch Universities of Applied Sciences Can Redesign Programming Assessment for AI
A strategic guide for Dutch universities of applied sciences redesigning programming assessment around validity, AI literacy, portfolios, oral evidence, process, workload, and policy consistency.
TraceYield Insights
13 min read · Updated September 18, 2026
TraceYield
AI engineering evidence
How Dutch Universities of Applied Sciences Can Redesign Programming Assessment for AI
This is a programme and institution question
A university of applied sciences cannot redesign programming assessment by changing one assignment in isolation. Students move between courses with different AI rules, lecturers share responsibility for learning outcomes, and examination boards need confidence in the validity and reliability of the programme as a whole.
The strategic question is how the institution prepares students for professional AI-supported work while preserving credible evidence of learning. That requires coordination across curriculum, assessment, policy, staff development, technology, and workload.
Review assessment validity first
Map what each assessment claims to measure, what students can outsource, and what evidence remains. Identify where a final take-home artifact is carrying too much of the judgment. Do not assume that every assessment must become supervised or oral; determine which outcomes need direct evidence and which can be assessed through authentic AI-permitted projects.
Validity should come before tool procurement. A platform cannot repair an assessment whose learning outcome, activity, and evidence are misaligned.
Build an institutional assessment mix
A coherent mix may include supervised fundamentals, AI-permitted projects, code reviews, portfolios, demonstrations, oral explanations, debugging tasks, and workplace-oriented deliverables. The mix can create both a no-AI lane and an AI-supported lane, each with a clear purpose.
Progression matters. Early assessments can establish foundations; later assessments can assess supervision, verification, security, and professional judgment. Students should understand how the expectations change and why.
Make policy consistent but adaptable
Institutional guidance should establish principles for approved tools, privacy, disclosure, data handling, academic integrity, and responsibilities. Course teams should translate those principles to their learning outcomes and assignment types. Examination boards need a process for reviewing exceptions and suspected breaches that does not rely on detector certainty.
Policy should be revisited when tools, regulations, or professional practice change. A static rulebook will quickly diverge from the work students and lecturers actually do.
Plan for lecturer workload
New evidence can create new grading work. Programmes should estimate the time needed for portfolios, oral explanations, moderation, and staff development before promising a format. Sampling, common prompts, shared rubrics, formative checkpoints, and selective evidence can reduce burden without removing judgment.
Technology may help organize or reconstruct available evidence, but it cannot eliminate the need for educational interpretation. Workload planning is part of assessment quality.
Include AI literacy as professional learning
Students should learn prompting and context specification, but also code reading, testing, security, licensing, privacy, critical evaluation, and knowing when not to use a tool. Lecturers need enough AI literacy to design tasks and interpret evidence. Leaders need a process for learning from pilots rather than announcing permanent solutions too early.
Npuls’ vision and related tools provide a national conversation starter. The programme must still translate those principles into its own outcomes, evidence, and governance.
A practical redesign sequence
Start with a small set of representative programming assessments. Map outcomes and risks, consult students and lecturers, identify the evidence gap, redesign one assessment mix, run it with moderation, and review the result. Record what became clearer, what became more burdensome, and which assumptions were wrong.
The objective is not a perfect one-time redesign. It is a coherent institutional capability to keep programming assessment valid as AI-supported development evolves.
What institutional leaders need to decide
Leaders should decide who owns the institutional principles, how programmes report assessment risks, what support lecturers receive, how students access approved tools, how privacy is handled, and how examination boards moderate new evidence. These are governance decisions as much as curriculum decisions.
A strategy that changes only the student-facing policy while leaving staff support, systems, and workload untouched will not remain coherent.
Measure whether redesign helped
Review whether students understand the permitted range, whether lecturers can assess evidence consistently, whether process and explanation improved validity, and whether workload is sustainable. Collect examples and limitations rather than a single “AI readiness” score. The institution should be willing to stop or revise an approach that creates more burden than educational value.
Redesign is a continuing quality cycle, not a one-off response to the arrival of a new tool.
Coordinate technology with pedagogy
An institution may be tempted to procure an AI-detection or activity platform before it has agreed what evidence its programmes need. The order should usually be reversed: clarify outcomes, map the assessment gap, define privacy and access requirements, then decide whether technology is necessary. A platform can organize evidence, but it cannot decide what learning means.
The same principle applies to approved coding agents. Tool procurement should follow a professional and educational use case, not create one by itself.
Make the strategy reviewable
Set a review period and collect examples of student work, lecturer workload, moderation findings, student questions, and examination-board decisions. Look for whether the new model improved the evidence of learning and whether it created unequal access or unnecessary documentation. Report limitations openly.
A reviewable strategy can evolve as evidence and tools change. A permanent rule announced without a feedback mechanism will become difficult to defend.
A leadership decision framework
Institutional leaders can ask five questions: what learning must remain directly visible; where can AI-supported professional practice be assessed; what evidence is proportionate; what support do staff and students need; and how will the institution review fairness and workload? The answers create a strategy that can be communicated and tested.
They also make it easier to explain why the institution did not adopt a single universal rule for every programming task.
Agent completion does not always mean engineering completion.
References
Pilot Program
Understand the WHY behind your engineering AI usage.
TraceYield evaluates trajectory evidence instead of stopping at spend totals. Join the private pilot to review AI coding usage with engineering context, security controls, and developer trust.
Join the TraceYield private pilot