Tertiary Education

Restructuring the degree

The three-year bachelor's degree is partly an artefact of the pre-digital era, designed around fixed bodies of knowledge that graduates would apply for a career. The challenge now is that knowledge updates faster than degrees do, and some specific skills are obsolete before graduation.

Several structural responses are being trialled internationally:

Shorter, modular degrees — two-year generalist foundations with optional specialisation stacks. Graduates enter the workforce earlier and return for further credentials as needs evolve. Proponents argue this matches the actual shape of AI-era careers; critics say it produces graduates too shallow for complex work and makes it harder for institutions to sustain research-linked depth.

Micro-credential stacking — building qualifications from verified short courses rather than monolithic degrees. New Zealand's NZQA framework already supports this. The question is whether employers, graduate schools, and immigration systems recognise micro-credentialed graduates equivalently — and whether they should. In most OECD countries, they do not yet.

Traditional degrees plus AI integration — keeping degree structures largely intact but embedding AI tools and AI-critical analysis throughout. This is the current de facto approach at most NZ universities, implemented unevenly and without a clear national framework. It requires less structural change but may underestimate how fundamentally AI alters what a degree should contain.

The NZ tertiary landscape — 8 universities, Te Pūkenga (now being restructured), and the wānanga sector — is large relative to the country's size. Coordination between institutions on degree structure is light-touch; the risk is that market pressures push towards homogenisation rather than useful diversity.

AI in teaching: from ban to embed

Every tertiary institution in New Zealand is currently managing a version of the same crisis: students are using AI writing tools, assessment systems weren't designed for this, and there is no consensus on what the right response is.

The institutional responses range from prohibition with detection software (treating AI use as academic misconduct and attempting to identify it through tools like Turnitin's AI classifier) to full integration (redesigning courses to assume AI availability and assessing what students do with it).

The prohibition approach has significant problems: AI detection software produces both false positives and false negatives, leading to unjust outcomes; students who comply are being trained in a world that their peers who don't comply are not; and the skill of working with AI tools is arguably one universities should be developing rather than suppressing.

Full integration requires rethinking what tertiary assessment is actually for. If a student can use AI to produce a first-rate essay, the question is what competencies an essay assessment was testing, and whether those competencies still matter. The honest answer depends heavily on the discipline.

Assessment reform is a live debate within NZ institutions. Some have moved rapidly; others are still in 2022-era policy. Government can either leave institutions to resolve this independently or establish a national framework — both positions have advocates.

Research priorities: where does NZ have an edge?

New Zealand cannot compete for dominance in foundational AI research. Training large models and publishing at the frontier requires capital and talent concentrations that small economies do not naturally produce.

The more interesting question is where NZ might have a genuine comparative advantage in applied AI research:

Primary industries — precision agriculture, fisheries monitoring, biosecurity threat detection, and conservation ecology are areas where NZ has world-class domain expertise combined with interesting applied AI problems. There is early evidence of genuine research quality here.

Indigenous language and knowledge — te reo Māori AI tools (speech recognition, translation, generation) require NZ-based research because the training data and the cultural context for evaluation both live here. This is an area where NZ is among the few places capable of doing the work, not merely one of many.

Small-country governance research — how AI regulation, labour market policy, and social systems adapt in small open economies is a research agenda that has value for many countries and that NZ is well-positioned to study through lived experience.

Where research investment should be concentrated — whether through Marsden, MBIE contestable funding, university QE category outcomes, or purpose-built national institutes — is itself a policy question with significant fiscal implications.

Wānanga: integration without erasure

New Zealand's three wānanga (Te Wānanga o Aotearoa, Te Wānanga o Raukawa, and Te Whare Wānanga o Awanuiārangi) operate under a kaupapa Māori framework. This creates a genuine tension that cannot be resolved by simply saying "AI is a tool and tools are neutral."

AI systems learn from data, and the data they have been trained on overwhelmingly reflects non-Māori knowledge systems, epistemologies, and worldviews. When a wānanga student uses a general AI assistant to support their learning, the embedded assumptions in that tool may actively conflict with what they are being taught.

The options include developing purpose-built te reo and tikanga-aware AI tools (expensive, technically demanding, but potentially transformative), establishing institutional norms that position AI as a subject of critical examination rather than uncritical adoption, or navigating the tension case-by-case at the course level. None of these is straightforward.

This is not a token inclusion — it is a genuinely unresolved policy and institutional question, and one where NZ has to lead because no other country has the same context.