Why this matters now
For much of modern education, producing an answer was evidence of having done a substantial part of the cognitive work required to reach it. Writing a coherent essay, creating a first design, structuring an argument or generating working code usually required a learner to move through many steps that were not visible in the final output.
Generative AI changes that relationship. A high-quality-looking output can now be produced with far less visible effort. This is not inherently a problem. It can improve access, speed up exploration and remove unnecessary mechanical work. But it creates an assessment and learning problem: the quality of the output no longer reliably tells us the quality of the thinking behind it.
Next Human Learning emerges from that gap. It asks what people still need to learn when intelligence is increasingly distributed across humans, machines, institutions and networks. The answer is not “less knowledge”. It is a richer relationship with knowledge: knowing enough to judge, knowing when to verify, recognising when an old model no longer fits, integrating different perspectives, deciding under uncertainty and learning from consequences.
This is a Mars Challenge conceptual position, not a claim that one educational model has been scientifically proven to be the correct response to AI. The available evidence supports the relevance of the problem and documents how Mars Challenge operationalises parts of the response; it does not yet validate Next Human Learning as a causal educational construct.
What Next Human Learning means
Next Human Learning names a shift from learning primarily as the acquisition and reproduction of answers toward learning as the capacity to keep rebuilding understanding in contact with reality.
The concept starts with an important refusal: AI capability does not make human knowledge obsolete. Without domain knowledge, people are less able to notice factual errors, missing context, weak assumptions or subtle category mistakes. Fluency can be mistaken for accuracy. Confidence can be mistaken for evidence. A tool can generate a sophisticated answer without knowing what matters in a specific social, ethical or institutional setting.
Next Human Learning therefore treats knowledge as infrastructure for judgement. People still need disciplinary depth. They also need enough intellectual range to recognise when a problem has crossed into another domain and when another person, discipline or source of knowledge needs to enter the decision.
Mars Challenge materials associate the concept with five elements: Learn, Unlearn, Relearn; Radical Creativity; Ethical Awareness; Collaboration & Resilience; Planetary Imagination. These should be read as an editorial architecture, not as validated psychological subscales. They describe the kind of learning the programme seeks to make visible.
The first element — learn, unlearn, relearn — makes learning dynamic. Learning builds models. Unlearning identifies when those models have become inadequate. Relearning reconstructs them using new evidence, context, experience or tools. The point is not perpetual novelty. It is the ability to keep knowledge responsive to reality.
Next Human Learning is also broader than AI literacy. Someone may be highly skilled at prompting, automation or model selection and still lack the capacity to assess evidence, understand trade-offs, work across disciplines or take responsibility for downstream effects. The central question is not “Can you use the machine?” It is “Can you remain a capable learner and responsible decision-maker while using it?”
From answer production to judgement
AI makes generation abundant. It can rapidly produce explanations, alternatives, drafts, images, code, strategies and synthetic viewpoints. But abundance of generation does not create abundance of judgement.
A plausible answer can still be wrong. A technically sound answer can be inappropriate to the context. An efficient solution can move costs elsewhere. A creative proposal can ignore implementation. A confident recommendation can conceal uncertainty.
This is where Next Human Learning shifts the educational objective. The learner must not only produce. The learner must interrogate what has been produced.
The Real-Life Readiness framework developed within the Mars Challenge ecosystem describes a closely related human burden in the age of generative AI: someone must still decide which question matters, what evidence is sufficient, what context is missing, which trade-offs are acceptable, what should become real and who is responsible for the result. That framing does not validate Next Human Learning, but it clarifies its purpose.
When generation is easy, educational design needs to make the hidden decisions visible. What sources were trusted? What remained uncertain? Which alternatives were rejected? What changed after feedback? What did the learner know independently, and where did they need external expertise? Why was this option chosen rather than another?
In this sense, Next Human Learning is as much about governing cognition as about acquiring it. It asks learners to understand not only an answer, but the conditions under which an answer should be trusted and acted upon.
Knowledge becomes more important, not less
AI access does not make domain knowledge less valuable. Prior knowledge enables people to recognise anomalies, ask better questions, detect dangerous simplifications and know when evidence is insufficient.
Next Human Learning does not require universal expertise. It requires enough depth to judge within a domain and enough breadth to recognise when a problem has crossed beyond it. Real problems routinely combine technical, behavioural, economic, ethical and institutional dimensions. Collaboration therefore matters intellectually: other people bring models of reality that an individual — or a generative system — may not see.
Learn, unlearn, relearn
The phrase is useful because it describes learning as a recurring relationship with change rather than a one-time transfer of information.
Learn: build enough structure to understand
Learning still means acquiring concepts, facts, methods, language and experience. But the goal is not maximal storage. It is building mental structure that can support interpretation and transfer.
A person has learned something deeply when they can recognise when it applies, adapt it to a new context and identify when it does not apply.
Unlearn: expose assumptions that no longer fit
Unlearning is not forgetting truth. It is withdrawing automatic trust from a model, habit or response whose conditions have changed. The more experienced a person becomes, the more difficult this can be: expertise creates useful patterns, but patterns can harden into blind spots.
Unlearning begins with the ability to say: “This worked before. I need to test whether it still works here.”
Relearn: rebuild competence, not just update information
Relearning can require reorganising a mental model, adopting a different method, integrating another discipline or changing how a tool is used. It is not a software update applied to an otherwise fixed learner. The learner changes with the context.
This matters in an AI-rich environment because roles, workflows and professional boundaries can change faster than formal curricula. A resilient learning system cannot predict every future skill. It needs to develop the capacity to reconstruct competence repeatedly.
The cycle is recursive. Action generates feedback. Feedback may expose a weak assumption. The assumption is revised. New knowledge enters. The next action becomes an experiment in the new model.
Four meta-capabilities that orient the cycle
Radical Creativity: going beyond the first plausible answer
When machines can generate many conventional options, creativity cannot simply mean producing more ideas. Radical Creativity is the capacity to reframe the problem, combine domains, explore improbable possibilities and resist premature convergence.
The Mars Challenge Participant Guide documents a process in which teams generate multiple ideas, discuss them and select a direction they can defend. That is operational evidence that divergence and choice are built into the programme. It is not evidence that every participant becomes more creative by a measurable amount.
Ethical Awareness: capability is not permission
A system may be technically feasible and still be unacceptable. Ethical Awareness places values, affected people and downstream consequences inside the learning process rather than after it.
This becomes especially important with AI. A model can recommend, rank or optimise, but it does not carry moral responsibility for implementation. Humans and institutions remain accountable for deciding how capability is used.
Collaboration & Resilience: learning through difference and revision
Collaboration in Next Human Learning is not merely efficient task allocation. It is a way to expose partial perspectives. Different disciplines, experiences and communities see different parts of a system.
Resilience adds the ability to continue learning when a hypothesis fails, feedback is uncomfortable or a chosen path must be abandoned. It is not endurance for its own sake. It is the capacity to remain adaptive without abandoning standards.
Planetary Imagination: expanding the boundary of the decision
Every action sits inside larger social and ecological systems. Planetary Imagination asks learners to consider resources, interdependence, long-term effects and what a solution shifts elsewhere.
This connects Next Human Learning directly to Earth-Back Thinking. Mars, as used in the Mars Challenge narrative, becomes an extreme context that makes constraints visible and then turns the learner back toward Earth with a sharper understanding of what is essential.
Productive friction is part of learning
Good tools remove unnecessary friction, but some friction contains the work we want learners to develop: choosing between incomplete options, defending a claim, verifying a source, responding to critique or admitting uncertainty. Next Human Learning therefore distinguishes productive friction from pointless difficulty. The aim is not to make tasks harder, but to preserve the cognitive operations that matter.
The Real-Life Readiness framework makes a compatible distinction: “real” does not mean unsafe. A learning environment can be bounded and supportive while still containing authentic users, constraints, trade-offs, feedback and consequences.
Mars Challenge provides operational examples. Teams move through structured phases, allocate responsibilities, make selections, construct prototypes, receive feedback and document the process. The Participant Guide asks them to record research, discarded ideas and changes, not only the polished final presentation. This design is compatible with deep learning because it makes revision and reasoning observable. The sources do not establish a causal effect, so that claim should remain a research question rather than a conclusion.
AI should amplify learning without erasing the learner
The most useful question is not whether learners should use AI. It is which cognitive work should remain visible and owned by the learner.
Four layers help make that distinction.
Generation. AI can expand the search space by producing alternatives, examples, drafts and simulations.
Verification. AI can assist with checking, but verification requires source quality, domain knowledge and contextual understanding.
Judgement. Choosing what the evidence means for this situation requires priorities, values and tolerance for incomplete information.
Responsibility. Someone must own the consequences of implementation. That cannot be outsourced simply because the recommendation came from a model.
The educational design question is therefore not “How much AI did the student use?” A more useful question is “Which parts of the reasoning did the student still have to understand, verify, choose and defend?”
A learner who uses AI to generate ten hypotheses and then investigates, rejects nine and defends one may exercise more judgement than someone who follows a manual procedure without AI. The technology alone does not determine learning quality. The architecture of the task does.
The Mars Challenge concept system
Next Human Learning becomes clearer when separated from three connected concepts.
Next Human Learning asks: How must human learning evolve?
Real-Life Readiness asks: Can a person mobilise what they know when reality is uncertain, interdisciplinary and consequential?
Agency asks: Can understanding become judgement, choice, action and responsibility?
Earth-Back Thinking asks: What happens when we use extreme constraints to expose what is essential and then bring that learning back to Earth?
Together they describe a coherent movement: keep learning and revising → meet reality → decide with authorship → act within limits and consequences. This is an editorial synthesis of the Mars Challenge knowledge architecture. It should not be presented as a validated developmental sequence.
What the evidence says
The available evidence has clear boundaries.
First-party concept definition. Mars Challenge 2026 programme materials explicitly define Next Human Learning as an evolution of education focused on meta-skills, creativity and ethical consciousness. They identify Learn/Unlearn/Relearn, Radical Creativity, Ethical Awareness, Collaboration & Resilience and Planetary Imagination as its central vocabulary. This verifies the concept’s institutional definition.
Adjacent first-party framework. Real-Life Readiness describes the human capacity to mobilise knowledge under uncertainty, distinguish evidence from plausibility, decide, act and take responsibility for consequences. It explicitly states that it is a foundational framework rather than a validated psychometric instrument and calls for future empirical work. The same evidence discipline should be applied to Next Human Learning.
External foundations already cited by the framework. The Real-Life Readiness document cites National Academies work on transfer, OECD work on agency and future education, UNESCO work on the futures of education, the World Economic Forum on changing skill demands, and recent research on generative AI and critical thinking. These sources support the wider problem space. They do not validate or endorse the proprietary Next Human Learning formulation.
Operational evidence. The Mars Challenge Participant Guide documents an environment requiring participants to make choices, take roles, research, prototype, receive feedback, revise and explain the Earth relevance of their work. This supports the claim that the programme creates conditions aligned with the concept. It does not establish that a specific learning outcome was caused by those conditions.
The appropriate evidence status is therefore Partial: the concept, rationale and implementation conditions are documented; specific causal efficacy remains unvalidated.
What we observe at Mars Challenge
The strongest first-party evidence available here concerns programme design and required behaviour, not measured psychological outcomes.
The Participant Guide requires teams to identify strengths, allocate roles, generate and select ideas, understand users, define problems, research, build, gather feedback, improve and document how a Mars-oriented solution returns value to Earth. It also treats the Concept Document as a record of process, including research, rejected ideas and learning, rather than only a final artefact.
This makes several elements of Next Human Learning observable: learners must move between knowing and not knowing; decisions cannot all be deferred to an instructor; feedback can require revision; different roles create different forms of accountability; and the final solution must be explained in relation to a wider system.
However, the documentation does not support a stronger claim that these conditions reliably produce durable changes in learning capacity. Demonstrating that would require measurement across time and contexts, including whether learners transfer the behaviour to new problems after the programme ends.
The distinction matters because Next Human Learning should itself model the epistemic discipline it advocates: do not confuse a plausible mechanism with proven impact.
Implications for learners, educators and institutions
For learners, the central shift is from proving independence from tools to proving authorship of judgement. Using assistance is not necessarily a weakness. The important questions are whether the learner can explain the problem, identify uncertainty, verify critical claims, recognise what came from elsewhere and defend the final decision.
For educators, task design becomes more important than tool policing. If an assignment can be completed entirely through automated production without any meaningful choice, it may no longer reveal the capability the institution cares about. Tasks can instead require evidence trails, oral defence, iterative prototyping, peer challenge, field context, reflection and transfer.
For schools and universities, disciplinary knowledge remains essential, but it needs opportunities to travel. Learners need contexts where knowledge crosses faculties, meets constraints and produces consequences. The institution does not need to abandon curriculum; it needs to create bridges between curriculum and reality.
For organisations, Next Human Learning suggests moving beyond one-off reskilling. Tools will continue to change. Sustainable capability lies in people who can reconstruct their skills, collaborate across functions, judge machine outputs and remain accountable for decisions.
Across all four groups, the underlying principle is the same: when intelligence becomes abundant, the scarce resource may increasingly be responsible human judgement capable of continued learning.
Related concepts
- Agency — the movement from understanding and judgement to choice, action and responsibility.
- Real-Life Readiness — integrated capacity to mobilise knowledge under uncertainty, interdependence and consequences.
- Earth-Back Thinking — designing under extreme constraints to rethink essential systems and return learning to Earth.
- Human Judgement — evaluating evidence, context, trade-offs and consequences before acting.
- Learning Transfer — applying knowledge beyond the setting in which it was originally acquired.
Evidence in practice
- Mars Challenge 2026 learning process — documented operational environment involving choices, roles, research, prototyping, feedback and revision.
- Mars Challenge Participant Guide 2026 — first-party documentation of the decisions, artefacts and reflection required across the learning journey.
No individual participant project is presented as causal proof of Next Human Learning because the provided sources do not include a project-level measurement of the concept.
Frequently asked questions
Does Next Human Learning mean subjects and disciplines matter less?
No. It assumes that judgement depends on knowledge. A person without sufficient domain understanding is less able to detect a plausible error, ask a high-quality question or know when specialist verification is needed. The change is not from “knowledge” to “skills”; it is from knowledge as something mainly reproduced to knowledge that can be transferred, questioned, combined and used responsibly.
Is Next Human Learning a validated learning theory?
Not in the materials provided. It is a defined Mars Challenge concept with documented components and an operational context. Related external research and institutional frameworks support the broader importance of transfer, agency, judgement, changing skill demands and responsible use of AI. But the specific construct has not been presented with validated measures or causal evidence. Its evidence status should therefore remain Partial.
What does “unlearning” mean here?
It does not mean discarding true knowledge. It means recognising when a habit, assumption, model or routine no longer fits the conditions. Unlearning is a form of revision: making previously invisible assumptions visible and withdrawing automatic trust from them. In rapidly changing environments, this can be as important as acquiring new information.
How should AI be used in Next Human Learning?
As a powerful learning and production partner without becoming the invisible owner of judgement. AI can generate, explain, simulate and help compare. The learner should still understand what matters, verify critical claims, recognise uncertainty, make choices and explain consequences. The right design depends on the learning objective: some tasks should be delegated; others should remain cognitively demanding because that demand is the capability being developed.
How is Next Human Learning different from lifelong learning?
Lifelong learning emphasises the need to continue learning over time. Next Human Learning adds a more specific response to an AI-rich and rapidly changing environment: continuous learning must include the ability to unlearn inadequate models, evaluate machine-generated outputs, transfer knowledge between contexts, collaborate across domains and retain responsibility for action.
How does it relate to Agency?
Next Human Learning focuses on keeping human learning adaptive and capable of judgement. Agency begins when that understanding becomes authorship: a person chooses, acts and accepts responsibility for consequences. In the Mars Challenge knowledge architecture, Next Human Learning prepares the learner to keep reconstructing competence; Agency describes what happens when that competence enters the world through responsible action.
References
Mars Challenge 2026 (Earth) — Program Framework / Manifesto Mars Challenge / Virtual Educa 2026 First-party source supplied in the project dossier.
Real-Life Readiness: A Foundational Framework Adelino Sousa / Mars Challenge / Virtual Educa 2026 First-party source supplied in the project dossier.
Participant Guide — Mars Challenge / The Grand Jam 2026 Mars Challenge 2026 First-party source supplied in the project dossier.
Education for Life and Work: Developing Transferable Knowledge and Skills in the 21st Century National Research Council / National Academies Press 2012 https://doi.org/10.17226/13398
OECD Learning Compass 2030 / Future of Education and Skills OECD External foundation cited in the Real-Life Readiness framework. https://www.oecd.org/en/data/tools/oecd-learning-compass-2030.html
Reimagining Our Futures Together: A New Social Contract for Education UNESCO 2021 https://unesdoc.unesco.org/ark:/48223/pf0000379707
The Future of Jobs Report 2025 World Economic Forum 2025 External foundation cited in the Real-Life Readiness framework.
The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects from a Survey of Knowledge Workers H.-P. H. Lee et al. / Microsoft Research & Carnegie Mellon University 2025 External foundation cited in the Real-Life Readiness framework.
Editorial note: the external references above support adjacent themes such as transfer, agency, changing skill demands and human-AI interaction. They should not be read as validation or endorsement of the proprietary Next Human Learning concept.
When AI generates answers, which human capabilities are still needed?
Generating a plausible answer does not remove human work. People still need to verify sources, recognise context, compare alternatives, decide under uncertainty, explain criteria, act and remain answerable for consequences. These capabilities do not establish abstract human superiority; they describe responsibilities the tool does not assume for a student.
An educational task can make them observable by asking for more than an output: the source of information, what was rejected, the criterion used, who may be affected and when the decision should be reviewed.
| Moment | Observable human action | Possible evidence |
|---|---|---|
| Verify | Confirm source, date and scope. | Verification and correction log. |
| Contextualise | Connect the answer with real people and constraints. | Context map and assumptions. |
| Decide | Select and justify an alternative. | Criteria, risks and decision. |
| Respond | Follow effects and revise. | Consequence and revision record. |
Hypothetical example: AI proposes three ways to save water. The team checks local data, consults users, rejects an unworkable option and documents when the chosen option will be reviewed.
Record AI use without confusing assistance with authorship
When AI use is permitted, transparency does not need to turn the work into an inventory of prompts. A useful record explains which task the tool supported, what it contributed, what the person verified or rejected and who made the final decision.
As an editorial practice, we propose a short assistance note. It should follow institutional rules, consent and data protection, and should not copy personal, confidential or irrelevant information. Declaring use does not prove authorship or learning, but it helps show where human verification, judgement and responsibility remained.
| Field | Question | Minimum record |
|---|---|---|
| Purpose | Which authorised task used AI? | Goal and limit of use. |
| Assistance | Which part came from the tool? | Type of support, without personal data. |
| Verification | What was confirmed, corrected or rejected? | Source, change and reason. |
| Decision | Who selected the final result and remains answerable for it? | Responsible person, criterion and next review. |
Hypothetical example: a team uses AI to suggest three presentation structures, rejects one because it omits risks, confirms two sources and writes the final version. The note documents support and decisions; it does not attribute the work to the tool or certify learning.
Explore next
- Agency — how understanding becomes responsible authorship.
- Real-Life Readiness — what it means to be prepared when reality has no answer key.
- Earth-Back Thinking — how extreme constraints can improve the questions we ask on Earth.
Next step
Deepen: Explore Agency
See evidence: See the Mars Challenge learning process in practice
Participate: Discover how to take part