Brazil
Planning + monitoring
AI supports organization and reflection, while child protection and equity constrain how learning evidence should be collected.
Students · Institutions · PolicyAIEOU COGNITION AND METACOGNITION SUBGROUP 3
A cross-country
framework &
toolkit.
What do metacognition and assessment redesign look like across countries and cultures in the age of AI?
Purpose: translate international evidence into a practical framework and toolkit for learning design, teaching, assessment, and implementation.
the learner
must stillthink
must stillthink
plan
monitor
evaluate
revise
Shared finding
AI can support learning only when design protects the learner’s active role in judgment, self-regulation, and reflection.
First: what differs?
Metacognition is not
the same intervention
everywhere.
All six contexts value planning, monitoring, evaluation, and reflection. What changes is the pressure point—the part of metacognition each context most urgently needs educational design to protect.
Switzerland
Judgment + justification
Metacognition becomes professional accountability: students surface assumptions, challenge their thinking, and defend decisions.
Students · Teachers · InstitutionsTürkiye
Verification + regulation
The emphasis is on resisting first-answer acceptance through comparison, evidence-seeking, ethical use, and self-regulation.
Students · Teachers · PolicyUnited States
Transfer + independence
Reflection must distinguish fast supported performance from durable mastery after AI scaffolding is removed.
Students · Teachers · InstitutionsUnited Kingdom
Process + shared language
A common taxonomy and visible evidence of revision help educators interpret and discuss metacognitive engagement consistently.
Students · Teachers · InstitutionsIndia
Meaning + conceptual connection
Bilingual, culturally familiar explanations connect prior knowledge to new concepts so students can monitor and explain understanding.
Students · TeachersThe difference that matters
Countries share a metacognitive goal, but they operationalize it through different evidence: protection, defense, verification, transfer, process, or meaningful connection.
Cross-country map
Six places.
Six pressure points.
Select a marker to move directly into that country’s Context, Challenge, Response, and Key insight.
Map base: BlankMap-World-v2, Wikimedia Commons (GNU Free Documentation License).
The comparative lens
One question.
Six distinct systems.
The countries differ in policy, infrastructure, educational level, discipline, and institutional readiness. Yet their direction converges.
Assessment is moving from “Can a student produce this?” toward “Can they explain, justify, evaluate, revise, and transfer the thinking behind it?” The website keeps national differences visible without flattening them into a universal solution.
Answer machineCompletionfluent product · hidden process
redesign
Thinking partnerInterrogationvisible judgment · accountable choice
Country perspectives
Same headings.
Different realities.
Select a country, then follow its context, challenge, response, and key insight. Audience tags show who each national picture concerns.
BR
Basic education · child protection · equity
Brazil
Policy momentum meets uneven, informal adoption.
StudentsTeachersInstitutionsGovernment policy
Context
National conversations emphasize responsible AI, human centrality, data protection, and use in basic education. In practice, access and guidance vary widely between schools.
Challenge
Students and teachers already use generative tools, often without curricular alignment or shared assessment rules. Restriction, quiet workload use, enthusiasm, and guilt coexist.
Response
Frame AI around planning, organizing, monitoring, and reflection—not answer production. Apply age-appropriate safeguards and treat analytics and detection as developmental supports, not surveillance.
Key insight
Metacognitive design and child protection belong together: learning evidence should make growth visible while minimizing data collection and preserving dignity.
Selected case studySwitzerland · teacher education
Decision-Centered Learning
Grade the judgment.
Not the artifact.
At the University of Teacher Education Lucerne, AI can help students build a didactic artifact. The artifact itself is not graded. The student’s reasoning and professional judgment are.
Ground
Develop an initial idea independently and anchor it in theory.
Student · TeacherChallenge
Use AI to question assumptions, compare options, and expose weak reasoning.
StudentDefend
Explain consequential choices with the artifact as a reference point.
Student · InstitutionWhy this case is strong It permits authentic AI use while preserving independent thought, productive challenge, and oral accountability in one aligned assessment sequence.
Whenever AI helps produce something, assessment can ask for the thinking behind it.
Then: what should change?
Recommendations follow
the comparison.
The cross-country evidence does not point to one universal AI rule. It points to a shared design logic that must be adapted by students, teachers, institutions, and government policy.
Who must act?
Responsibility is distributed.
Interrogate
Plan, question, compare evidence, explain accepted and rejected outputs, and transfer learning without AI.
Design
Protect productive struggle, set transparent boundaries, and assess reasoning, revision, judgment, and reflection.
Enable
Build shared language, professional learning, privacy safeguards, and consistent policy-to-practice support.
Govern
Set rights-based, age-appropriate, equitable expectations while preserving teacher judgment and public trust.
A practical design sequence
Start with learning.
Then decide on AI.
The report’s framework shifts the central question from whether AI was used to how AI use affected learning.
- 01
Learning goal
What should learners know or be able to do?
- 02
Cognitive task
What can AI do—and what might be offloaded?
- 03
Metacognitive task
What planning, monitoring, and judgment stays human?
- 04
Evidence
How will the learner’s thinking become visible?
- 05
Assessment
Can they explain, defend, evaluate, and transfer?
- 06
Design
Where should we preserve struggle, agency, and choice?
Desired evidence
Prompts and logs are not the goal. What matters is credible evidence of reasoning, judgment, revision, reflection, and transfer.
Across all six contexts
Five signals for redesign
Validity
A final product cannot be assumed to represent independent understanding.
Visibility
Capture planning, comparison, monitoring, revision, and self-explanation.
Interrogation
Use AI to question and test ideas—not merely to complete tasks.
Teacher judgment
Technical fluency cannot replace pedagogical and cultural discernment.
Metacognition
Learners must notice, regulate, and explain the thinking AI changes.
Consolidated reference list
Sources cited in the report
Full references have been moved here and numbered for unobtrusive in-text linking. The OECD entry remains identified as a working reference because that is how it appears in the supplied document.
- 01
Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), e2422633122. Open source ↗
- 02
Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., Shen, Y., Li, X., & Gašević, D. (2024). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56(2), 489–530. Open source ↗
- 03
Mishra, P., & Koehler, M. J. (2006). Technological pedagogical content knowledge: A framework for teacher knowledge. Teachers College Record, 108(6), 1017–1054. Open source ↗
- 04
OECD. (2024). Artificial intelligence and the future of skills and education. [Working reference as supplied in the report.]
- 05
UNESCO. (2023). Guidance for generative AI in education and research. Open source ↗
- 06
Wing, J. M. (2006). Computational thinking. Communications of the ACM, 49(3), 33–35. Open source ↗
- 07
Menary, R. (2007). Writing as thinking. Language Sciences, 29(5), 621–632. Open source ↗
- 08
Swiss Federal Council. (2025). AI regulation: Federal Council to ratify Council of Europe Convention. Open source ↗
- 09
swissuniversities. (2024). Artificial intelligence in university teaching. Open source ↗