Key Takeaways

• Learning exists to reduce uncertainty, both for individuals and for humanity as a whole.
• Education should focus less on prediction and more on building adaptable judgement, methods, and values.
• Learners, parents, and tutors can plan for uncertainty by building robust capability portfolios rather than narrow career bets.

Learning as uncertainty reduction

In October 2025, three of us were in Nettlecombe, Somerset, with undergraduate environmental science students. We were staying in a twelfth-century building, surrounded by a landscape that, in its essentials, has remained recognisable for centuries.

One evening, the discussion turned philosophical. If artificial intelligence continues to advance, and if one day machines and robots can perform almost every task humans currently do, what is the objective of learning? Why do humans need to learn at all?

The setting mattered. People who lived in that building eight hundred years ago learned. We learn now. People in the future will learn. The tools change. The reason does not.

The question appears everywhere today: why learn if artificial intelligence can already do so much of the work?

The question sounds practical, although it rests on a narrow assumption. It treats learning as preparation for a job, rather than as a fundamental human process. I see learning differently. Learning exists to reduce uncertainty. This is true at the level of the individual, and it is true at the level of humanity.

When I use the term uncertainty, I mean the lived form of human needs. A need appears to us as uncertainty about how to act, what to trust, or what will happen next. Curiosity and play do not sit outside uncertainty reduction. They are how humans are drawn toward new uncertainty rather than away from it, seeking exploration, wonder, and meaning.

An individual learns to understand the world well enough to act. A society learns, through research and shared institutions, to reduce uncertainty about nature, health, technology, and risk. This process has always been unfinished. Every solution exposes new questions. Progress moves the boundary of uncertainty, rather than eliminating it.

Artificial intelligence does not change this basic reality. It changes the speed.

Individual learning: reducing personal uncertainty

To experience life, a person must reduce uncertainty.

Human experience depends on orientation. We need to know enough about the world to move through it, make choices, and live with the consequences. This is why learning begins long before careers enter the picture.

At the individual level, learning reduces uncertainty about what is happening, what matters, and what actions are sensible.

Often, the knowledge needed to reduce this uncertainty already exists at the level of humanity. It may be stabilised in books, embodied in experienced people, or embedded in institutions and systems. The uncertainty persists because the individual does not yet possess that knowledge or skill.

Individual learning is therefore not the creation of new knowledge, but the transfer, interpretation, and internalisation of existing human understanding.

A learner builds mental models. These models explain how things work well enough to make decisions. Writing clearly, interpreting numbers, understanding cause and effect, and learning how to check claims are all forms of uncertainty reduction.

This is why learning is not exhausted by access to information. Information without understanding does not reduce uncertainty. Understanding requires practice, feedback, and revision. It requires effort. Artificial intelligence can assist with practice and feedback, but it cannot decide which models a learner should trust. The greatest risk is not error, but effortlessness. Learning that leaves no cognitive scar tissue rarely produces judgement.

Learning how to learn is therefore not optional. Methods outlast content.

Most early learning has little to do with jobs. Children do not learn to speak, read, count, or reason in order to enter a labour market. They learn to function, to communicate, and to make sense of their experience.

Only later do societies channel learning towards work. At that point, individuals specialise. They learn particular skills so they can contribute to society and, in return, receive resources that allow them to access other services.

This uncertainty is therefore not primarily about the absence of jobs. The deeper structure is more fundamental. Human life requires services to be provided. Services exist because people need uncertainties reduced in order to experience their lives meaningfully.

Jobs are one historical way of organising who reduces which uncertainties. They are downstream of uncertainty reduction, not its source.

Jobs are therefore an organisational layer, not the purpose of learning. They are one way societies decide who reduces which uncertainties. When conditions are stable, this mapping feels predictable. When conditions change, the mapping becomes fragile.

The durable asset is not a specific job skill, but the capacity to update one’s understanding and contribution as uncertainties move.

Collective learning: reducing humanity’s shared uncertainty

Some uncertainties cannot be resolved by individuals alone.

Beyond what can be learned from books, mentors, or existing systems, there are questions that require collective effort over time.

At the level of humanity, learning appears as research, experimentation, and institution building.

Science reduces uncertainty about natural systems. Public health reduces uncertainty about exposure and risk. Engineering reduces uncertainty about whether systems will fail. Social science reduces uncertainty about how people and organisations behave.

This collective learning feeds back into individual lives. Clean water systems, vaccines, and communication networks are the result of accumulated uncertainty reduction across generations.

Artificial intelligence enters this picture as a powerful accelerator.

What was once humanity’s frontier uncertainty often becomes, over time, individual uncertainty. Once research stabilises knowledge, it enters education. What was unknown becomes teachable. What was risky becomes routine.

This cycle has repeated throughout history. It allows faster exploration of ideas, faster testing of hypotheses, and faster dissemination of results. However, it also creates new uncertainties, including ethical risk, bias, misuse, and over-reliance. Collective learning must now operate at a higher speed, with greater responsibility.

Why AI increases the need to learn

Artificial intelligence compresses feedback loops.

For most of human history, learning required direct engagement with the world. People went into nature to observe, to test, and to fail. Feedback came slowly, from seasons, from success and loss, from cause and consequence.

Later, feedback increasingly came from other humans. Knowledge passed from one generation to the next through apprenticeship, oral tradition, and teaching. With writing and books, learning accelerated again. Feedback could travel across space and time.

The modern period added formal research, institutions, and then digital networks. Websites and databases reduced the time needed to access accumulated human knowledge, although interpretation still required effort.

Artificial intelligence marks another compression. Drafts are produced faster. Plans are generated instantly. Answers appear fluent and confident.

What once took years, then months, then weeks, can now happen in moments.

The main shift introduced by artificial intelligence is not the removal of uncertainty, but the relocation of uncertainty. This is why certain jobs become fragile. Those jobs existed because they were efficient ways of allocating particular uncertainty-reduction tasks to humans.

Artificial intelligence now allows a single person, augmented by tools, to resolve uncertainties that previously required many. This can temporarily disrupt familiar forms of work and service delivery.

Uncertainty moves:

• from execution to judgement
• from performing tasks to deciding which tasks matter
• from producing answers to evaluating their consequences

As uncertainty moves, needs reconfigure learning becomes more important, not less. Learners must develop the capacity to recognise where uncertainty now sits, and how to respond to it responsibly through flexible, adaptive learning.

The real educational challenge

Learner uncertainty today is therefore not a sign that learning has lost its purpose. It is a signal that prediction-based education has reached its limit.

Historically, education worked because there was a relatively stable mapping:

learn knowledge → acquire skills → perform a role → contribute → earn → live

Artificial intelligence disrupts this predictability. Not because contribution disappears, but because its future location becomes harder to foresee.

Learners feel uncertain because they cannot predict which uncertainties society will ask them to resolve when they finish their education. This uncertainty is rational.

The mistake is to interpret this as a crisis of learning. It is not. It is a crisis of over-reliance on forecasting.

The deeper reality is this. Human life will always require uncertainty reduction in order to be experienced meaningfully. Artificial intelligence resolves some uncertainties faster, but it also creates new ones. The overall demand for judgement, responsibility, and sense-making does not decline.

Education systems that over-optimise for narrow predictions become fragile. Education that builds adaptable judgement remains robust.

The issue is not that uncertainty will disappear. The issue is that uncertainty will move, and learning must move with it.

Adaptive Learning as Uncertainty Reduction (ALUR framework)

If learning is uncertainty reduction, then education should prioritise capabilities that remain useful when contexts change.

Learners need:

• models that explain how the world works
• methods for investigating claims and evidence
• values that guide priorities and responsibility
• experiences that turn knowledge into judgement

These elements support both individual agency and collective progress. We proposed a framework Adaptive Learning as Uncertainty Reduction (ALUR) to allow learners to act without pretending to certainty.

ALUR is not a replacement for existing educational taxonomies. It sits alongside them, in the same way that Bloom’s taxonomy clarifies cognitive depth. Where Bloom focuses on levels of thinking, ALUR focuses on where uncertainty sits, and how learners adapt as it moves.

ALUR exists because prediction-based education is reaching its limit. In a world shaped by artificial intelligence, adaptability-based learning becomes more valuable.

The framework is organised in layers rather than paths. Learners develop all layers in parallel, with emphasis shifting over time.

A critical assumption sits underneath ALUR. The unit of uncertainty reduction is shifting.

Historically, uncertainty was reduced mainly through services designed for groups. Jobs existed to deliver those services at scale, and individuals fitted themselves into those pipelines. One teacher served many students. One doctor served many patients. One product or software system served an organisation or a market.

Artificial intelligence changes the economics of personalisation. It becomes feasible to tailor responses, tools, and services to individuals. As a result, uncertainty reduction increasingly happens at the level of the individual, not only through mass services.

Human contribution remains human-to-human. What changes are the form, the means, and the scale at which uncertainty is reduced. The future of work increasingly resembles a series of micro-contributions, where individuals repeatedly reduce uncertainty for specific people, rather than a single role held over a forty-year career.

Layer 1: Foundations for orientation

This layer supports the most basic form of uncertainty reduction: understanding enough to function.

Learners develop:

• deep reading and listening
• clear writing and communication
• numeracy and quantitative intuition
• basic scientific and causal reasoning
• attention, focus, and self-regulation

These foundations allow learners to interpret the world rather than react blindly to it. Without this layer, higher learning collapses into surface fluency.

Layer 2: Methods for finding out

When information is abundant, uncertainty shifts from access to evaluation.

This layer focuses on how learners reduce uncertainty when answers are incomplete, contested, or misleading.

Learners practise:

• asking good questions
• searching and filtering information
• comparing sources
• spotting assumptions
• verifying claims

Artificial intelligence belongs here as an enabler. It accelerates exploration and comparison, but learners remain responsible for judgement.

Layer 3: Judgement, values, and responsibility

Not all uncertainty is factual. Much of it concerns priorities, trade-offs, and consequences.

This layer develops the capacity to decide what matters, and to act responsibly under imperfect knowledge.

Learners engage with:

• ethical reasoning
• conflicting viewpoints
• justification of decisions
• revision of beliefs when evidence changes

As automation increases, this layer grows in importance. It is where human agency remains central.

Layer 4: Making and delivering uncertainty reduction (self to others)

This layer converts learning into contribution.

Learners identify real uncertainties and design responses that reduce them. The focus progresses over time:

• early stage: reducing uncertainty for oneself
• later stage: reducing uncertainty for others

This may take the form of:

• tools, guides, or workflows
• explanations or teaching materials
• digital or physical artefacts
• services delivered to specific people

Artificial intelligence accelerates prototyping, iteration, and feedback. Learners remain accountable for whether uncertainty actually reduces.

A practical test for this layer is simple:

Can the learner point to something they made that reduced uncertainty for a specific person? This also implies a shift in assessment, from testing recall to verifying impact.

How ALUR scales with age and education

In early education, emphasis naturally sits on Layers 1 and 2, with guided exposure to Layer 3.

In secondary education, learners begin to integrate judgement and small-scale projects.

In higher education and beyond, Layer 4 becomes central. Learners are expected to demonstrate contribution, not just comprehension.

Why ALUR matters in the AI era

Artificial intelligence resolves some uncertainties faster than humans ever could. At the same time, it makes the location of future uncertainty harder to predict.

ALUR provides a way to prepare learners without forecasting specific jobs. It focuses on building people who can adapt as uncertainty moves.

Learning remains essential, not because work exists, but because uncertainty does.

---

This article was created with the assistance of AI tools, including ChatGPT, and Google AI, for research, structuring, and image generation.