Rebalancing the social contract of higher education in the age of AI

When the degree stops being scarce

AI does not remove the need for higher education, but it does change the basis of its value. Scarcity is shifting from access to information towards trust in judgement, and towards performance under real constraints.

Scarcity moves from content to credibility

In many sectors, AI is already compressing the time between novice and competent contributor. Junior software developers can generate working code with assistance, early career analysts can draft memos quickly, and marketing teams can prototype campaigns at speed. This does not eliminate expertise, but it changes the meaning of entry-level competence. If baseline production becomes abundant, the differentiator becomes the ability to frame problems, evaluate trade-offs, spot failure modes, and act responsibly when the model is confident but wrong.

The social contract of higher education has historically rested on credible scarcity: selective admissions, limited instructional bandwidth, and assessment regimes that separate mastery from participation. AI reduces some of that scarcity, including the scarcity embedded in writing, summarising, translating, and drafting. If those outputs are no longer reliable evidence of student capability, the degree risks becoming a weaker proxy for what employers and society actually need.

Capability becomes situational, not generic

Organisations increasingly hire for performance in specific contexts, not just general intelligence. In regulated environments such as financial services, pharmaceuticals, and critical infrastructure, the question is not whether someone can produce an answer, but whether they can justify decisions, document reasoning, and operate within controls. AI makes “knowing” easier, and makes “knowing when not to trust” more valuable. The institutional challenge is to evidence that distinction without turning education into mere compliance.

Public subsidy, private capture

The AI era sharpens an old question: who benefits from public investment in higher education. If AI disproportionately increases returns for those already advantaged, the legitimacy of funding models comes under pressure.

Graduate advantage may decouple from learning

In many systems, the degree has operated as both a development experience and a sorting mechanism. AI introduces a risk that sorting intensifies while development becomes harder to evidence. If high-status institutions can use brand, networks, and selective pipelines to maintain wage premia, while AI erodes the uniqueness of what is taught, public confidence can weaken. This is not a claim that prestige is undeserved, but a prompt to ask what portion of advantage is linked to demonstrated capability versus inherited signalling power.

There is also a distributional question. AI tools are already unevenly adopted across households, employers, and regions. If some learners arrive with better AI fluency, better devices, more time, or more supportive workplaces, the productivity uplift can compound existing inequality. In that scenario, higher education can accidentally become a multiplier of private advantage, even when intentions remain inclusive.

Research, data, and intellectual property add new asymmetries

As universities partner with AI vendors, new forms of private capture can emerge. Training data, student interaction data, and curriculum artefacts can become assets. When platform providers gain privileged access, the public funding that supported knowledge creation may indirectly underwrite private model improvement. The question is not whether partnerships are wrong, but how the terms preserve public value, including transparency, accountability, and reciprocal benefit.

A practical governance dilemma follows: are AI systems being procured as tools that serve education, or are institutions drifting into dependency where education serves the tool’s data needs?

Assessment becomes the new currency

If AI can produce plausible assignments, assessment must evolve. The goal is not to outsmart AI, but to measure what matters: judgement, integrity, and real-world performance.

From artefacts to decision trails

Traditional assessment often privileges polished artefacts: essays, reports, take-home projects. AI makes artefacts cheap. A more resilient approach treats the artefact as a by-product and evaluates the decision trail: what was assumed, what evidence was selected, what alternatives were rejected, what risks were acknowledged, and what would change the conclusion. In corporate settings, strong governance increasingly demands this. Audit committees, clinical governance boards, and safety regulators care less about rhetorical fluency and more about explainable choices.

Some professional firms have begun testing candidates through supervised case simulations, live problem-framing, and collaborative exercises where reasoning is visible. Higher education has an opportunity to learn from that shift rather than defend assessment formats whose evidential value is declining.

Simulations and performance tasks scale differently with AI

AI can also expand assessment possibilities. Scenario-based role play can test ethical judgement, stakeholder management, and decision-making under pressure. Repeatable simulations can allow consistent standards while reducing reliance on one-off written work. At LSI, for example, AI-supported simulations and formative feedback are being used to make mastery more observable, with human sessions focused on sense-making rather than marking volume. This points to a broader possibility: AI as a tool to increase rigour and feedback density, not to lower standards.

An empirical study that could reduce uncertainty

A sector-wide longitudinal study could track graduates assessed through AI-enabled performance tasks versus traditional written assessments, measuring subsequent workplace outcomes such as time-to-autonomy, error rates in controlled tasks, progression speed, and supervisor confidence. If designed with employer partners, independent oversight, and privacy safeguards, such evidence could move debate beyond anecdotes about “AI cheating” and towards verifiable claims about capability.

Institutional purpose under AI pressure

AI makes it tempting to reduce higher education to content delivery. Yet the enduring value may sit in curation, community, and the legitimacy of standards. The question becomes what institutions are actually for, when information is everywhere.

Universities as trust infrastructure

When AI makes it easier to generate convincing but unreliable outputs, trust becomes an infrastructure problem. Higher education can contribute by setting norms for evidence, reasoning, and responsible use of tools. This resembles the role of accounting standards in markets: not exciting, but foundational. The more AI pervades society, the more valuable credible standards become, provided they remain connected to real performance rather than ritual.

Teaching models shift from delivery to continuous improvement

AI also challenges the operating model. Term-based delivery, fixed syllabi, and episodic assessment were shaped by physical constraints. AI reduces those constraints and exposes new expectations: fast feedback, adaptive pathways, and ongoing skills refresh. Some institutions will treat this as an efficiency play, cutting costs by automating interaction. Others may treat it as an opportunity to reallocate human expertise towards mentoring, judgement coaching, and high-stakes evaluation.

Different futures remain plausible. If AI drives down the cost of basic instruction, some systems may widen access at lower price points. Alternatively, premium models may double down on intensive human experience and social capital. Resilience may come from designing institutions that can operate credibly in either scenario.

Governance for a redesigned bargain

Rebalancing the social contract is partly a design question and partly a governance question. Decisions about data, procurement, quality assurance and funding will shape whether AI strengthens public value or accelerates private capture.

Procurement choices become educational policy

Vendor selection is no longer a technical matter. Contract terms influence academic freedom, student privacy, assessment integrity, and long-term cost. Useful questions include whether models can be audited, whether student data is used for training, what happens if a provider changes pricing, and how easily an institution can switch platforms without losing core learning records.

Quality regimes need a theory of AI

Regulators and accrediting bodies face a delicate task: encouraging innovation while protecting students and public confidence. Overly rigid rules can freeze improvement; overly permissive rules can allow a race to the bottom where credentials detach from capability. A pragmatic direction is to require evidence of learning validity under AI conditions, rather than mandating specific pedagogies. That shifts attention from inputs to outcomes, while still recognising that outcomes are complex and take time to observe.

Funding models may need new fairness tests

If AI changes who benefits from higher education, public funding may demand clearer lines of sight to public value. This does not have to mean crude short-term employability metrics. It could mean demonstrable contribution to local innovation ecosystems, productivity diffusion into SMEs, capability-building in public services, or measured improvements in social mobility. The hardest part is agreeing what to measure, and what trade-offs society is willing to accept.

Decision test: would the institution make the same AI choices if graduate wage premia weakened, but civic and workforce needs intensified?

Uncomfortable question: if the credential no longer guarantees private advantage, is the institution still confident it is delivering public good that society can recognise and defend?