U.S. Institute of Public Education
Parent, educator, and adult learning
AI Explorers
I think. I choose. I create.
Learning with purpose
An applied learning programme from the U.S. Institute of Public Education. Build the judgment to use transformative AI tools in teaching, learning, and life transitions—with human agency at the centre.
Start here · 5 minutes
Help the learner remain the author
Imagine a child submits an attractive playground design. An AI tool awards it a high score. The useful next conversation asks how the child chose its features, whose needs were considered, what evidence supports the design, and which parts still need checking. A score opens questions; the adult’s job is to help make the learner’s thinking visible.
What can the learner explain, choose, check, and use in a new situation after receiving help?
The activities use fictional information. Paper, conversation, and a child’s existing AI Explorers workbook are enough. No child AI account is needed.
By the end, you can
- State the learning goal before choosing an AI tool.
- Recognise inaccurate or unhelpful feedback and replace it with a useful next step.
- Observe understanding through explanation and a new example while keeping access supports available.
- Set clear responsibilities for privacy, permission, review, and human support.
- Offer several ways to show the same learning and respond respectfully when a child questions feedback.
Follow the five lessons for eight minutes each, spend ten minutes on the shared case, then five minutes on your plan. You may also use five short meetings. Revisit a new example after a few days.
Lesson 1 · 8 minutes · AI Can Help
Understand the tool
Identify the tool’s purpose, its limits, and the adult who checks its output.
Core concept
A fluent response can contain mistakes, unsupported assumptions, or advice that misses the learning goal. Begin with the actual task. Identify whether the tool suggests ideas, gives feedback, or makes a judgment. Check its output against the question and available evidence before sharing it with a child.
Feedback laboratory
A fictional child writes 12 ÷ 3 = 5. Compare three fictional comments: A says, “Five is correct.” B says, “The answer is four.” C says, “Arrange twelve counters into three equal groups. How many are in each group? Explain your picture.” Identify the error in A, the limited teaching value in B, and the reasoning invited by C. Rewrite one comment to include an accurate next step.
Consider a teaching response
An effective revision is: “Try twelve counters in three equal groups. Count each group, then show why your answer fits.” Offer a worked example when the child needs instruction to begin; adjust the amount of help to what the child currently understands.
At home
Sit alongside the child. Check the suggestion before using it. Ask which part helps and which part feels confusing.
“Let’s check this together. Show me where the suggestion fits your work.”
Reflect and apply
Lesson 2 · 8 minutes · My Ideas Matter
Make thinking visible
Distinguish an assisted product from evidence of understanding.
Core concept
Choose one specific learning goal, such as explaining equal groups or checking a design against two criteria. Invite a first idea before presenting suggestions. After help, ask the learner to explain one choice and try a similar example. Keep ordinary access supports, such as counters, drawing, speech, translation, or assistive communication, available.
Same idea in a new setting
Plan three moments: the child groups twelve counters into three equal groups with support; the child explains the picture in their chosen format; later, the child considers fifteen counters in three equal groups. During the later example, avoid AI supplying the answer. Record the support used and what the child explains.
Consider a teaching response
The goal is an explanation of equal groups. A correct answer with an unclear explanation invites further teaching. A thoughtful explanation using counters can show understanding even when writing is difficult. Use observations to choose the next lesson; allow a learner to ask for information or more practice.
At home
Ask a short explanation after a joint activity. Revisit the idea during a different everyday example.
“Which idea came from you? Why did you keep or change it? Where else could it work?”
Reflect and apply
Lesson 3 · 8 minutes · Ask, Check, and Get Support
Share responsibility
Define permissions, decision roles, and a reliable route to human review.
Core concept
Adults set the purpose and limits of a learning activity. Explain what information may be used, who checks advice, when the activity pauses, and how a child can ask another person for help. A child’s question about feedback deserves examination. Decisions that affect grading, placement, or opportunity require accountable human judgment and a clear way to request review.
A child challenges the feedback
A fictional AI comment says a drawing has no quiet corner. The child points to a labelled quiet corner. Role play the child, the adult, and the human reviewer. Inspect the drawing, identify the mistake, and decide how the feedback will be corrected. Let the child explain whether any revision is needed.
Consider a teaching response
A useful response is: “Your drawing shows the quiet corner. The comment missed it. Let’s correct the feedback and check whether any other point needs attention.” The adult verifies the evidence and follows the setting’s review procedure.
At home
Agree on one adult role for help and a second role when the first person is unavailable. Use fictional examples to practise asking.
“You can question this result. Let’s look at the evidence and find the person who can resolve it.”
Reflect and apply
Lesson 4 · 8 minutes · AI Can Help
Give useful feedback
Choose feedback that is accurate, understandable, and actionable.
Core concept
Start by recognising what the learner is trying to do. Compare the work with the learning goal. Give one manageable next step, then invite a response. Questions, demonstrations, explanations, and worked examples each have a place. The adult selects the kind of support that helps this learner move forward.
Rewrite a vague comment
A fictional tool comments, “Improve your playground.” Rewrite it using the agreed goal: include a quiet area and a route without steps. Name an observed feature, identify a specific gap, and offer one next action. Then ask the child to explain the change they choose.
Consider a teaching response
“Your plan includes a quiet area. The entrance route is still unclear. Draw or describe how someone could reach the quiet area without steps, then explain your route.” This feedback points to evidence and leaves a meaningful design decision with the learner.
At home
Use one question at a time. Pause when frustration rises and return with a demonstration or simpler example.
“Here is what your work already shows. Choose one next step and tell me your reason.”
Reflect and apply
Lesson 5 · 8 minutes · My Information Matters
Protect participation
Protect information while offering fair ways to participate and show learning.
Core concept
Design the activity so it works with invented information and without a child’s own AI account. Adults check any tool’s age terms, data practices, and suitability before use. Use only the information required for the learning purpose. Offer speech, drawing, writing, or building when these can demonstrate the same goal. Access support stays available during learning checks.
Privacy editing studio
Edit this fictional prompt: “Write feedback for Maya, age eight, who attends North School, has a medical diagnosis, and lives at 12 Cedar Road. Her playground drawing needs a quiet area.” Replace it with: “Suggest one clear question for a fictional playground design that needs a quiet area.” Explain what the task actually needs. Then design a spoken and a drawn response to the same goal.
Consider a teaching response
The task needs a design goal and a generic description of the work. Personal identifiers and a medical history add unnecessary disclosure. Equivalent response choices can make the learner’s ideas visible while the adult applies the same learning criteria.
At home
Use invented details. Ask before sharing another person’s work or image, then check whether sharing is appropriate.
“Your information matters. You can show your understanding in a way that works for you.”
Reflect and apply
Research Update · 23 October 2026
Keep Your Ideas. Check Your Choices.
Five short activities help children ages 7 to 10 explore learning from examples, preserve their ideas, check evidence, protect information, and explain choices through the playground project. Choose one with a parent or educator. Paper, drawing, and conversation are enough; no child AI account is needed.
Optional extension: allow 5 to 10 minutes for one activity, with a brief revisit after a few days. This is additional to the 60 minute core course.
Before starting, name one learning goal and agree on a pause signal. Invite a first attempt before offering a hint. Keep ordinary access supports available: reading aloud, translation, drawing, counters, or an adult recording the child’s words. Ask what helped, then adjust the next prompt. Participation may pause or end at the child’s request.
Activity 1
Train, Test, Try Again
Meet the Pattern Finder · A Playground Where Everyone Belongs
Choose one visible feature for sorting fictional playground picture cards, such as whether a bench has a backrest. Give examples to an adult playing the Pattern Finder. Keep two cards hidden for testing. Try an unfamiliar card, inspect a mistake, then revise the examples.
Ask the Learner
What pattern did the finder learn? Which examples were missing? What can this feature tell us, and what can it not tell us?
Facilitator Guidance
This is a paper analogy for learning from examples, not an actual trained AI model. A rule followed by an adult is not machine learning. One visible feature does not establish that a playground suits everyone; children choose what matters and identify further questions.
Guide the Practice
- Let the child choose the feature and label several fictional picture cards. Set aside two unseen examples. Use drawings of objects, not images of people.
- The adult infers a pattern from the labelled examples and predicts a hidden card. The child compares the prediction with the chosen feature. Include a different-looking example or a card whose feature is unclear.
- Discuss a mismatch or uncertainty, add a useful example, and test the other hidden card. Stop after one revision. Ask whose needs the sorting feature leaves unanswered.
A Possible Learner Response
“It guessed that every long seat has a backrest because our examples looked like that. I will add a short bench with a backrest. A backrest alone does not tell us whether everyone can reach the bench.”
If More Support Is Needed
Offer two visible features to choose between, read cards aloud, or use tactile object models. Keep the child’s feature choice. Explicitly distinguish a paper simulation from a machine that adjusts a model using data.
What to Notice
- Chooses a feature and gives labelled examples.
- Checks a prediction using a previously unseen card.
- Explains one limitation or a useful revision.
Try a New Situation
On another day, sort fictional library furniture using a new feature. Ask the learner to choose an unseen test example and explain why more examples may help without guaranteeing a correct prediction.
Learning area: AI Can Help · My Ideas Matter
Activity 2
Start With Your Own Idea
My Playground, My Ideas · Ask, Check, Improve
Draw a playground before seeing suggestions. Complete the three-box reasoning card: My idea; Evidence I need; What remains unknown. Keep the first drawing. An adult reads: “A bigger slide would make your playground better.” Use your goal and evidence to keep, change, or decline the suggestion.
Ask the Learner
Does this suggestion serve my goal? What might improve, and what might we lose?
Facilitator Guidance
Look for a child-owned priority and a reason connected to it. Agreeing with AI is not the success criterion. Declining a suggestion can show sound judgment; changing an idea can also be thoughtful when supported by a reason.
Guide the Practice
- Name or show one priority. Record it in My idea before revealing any suggestion.
- In Evidence I need, choose a fact to check in the fictional plan. In What remains unknown, name a question the plan does not answer. Ask an adult to help check.
- Read one fictional suggestion. Compare it with the original goal and checked evidence. Keep, change, or decline it, explain why, then stop at the agreed limit.
A Possible Learner Response
“My idea is a quiet place. I need to check whether it has shade. The plan says it has none. I still do not know how wide the other routes are. I will keep the quiet corner and consider adding shade.” This is an example, not an answer to copy.
If More Support Is Needed
Use sketches, spoken responses, movable cards, or adult scribing. Ask one box at a time. Do not require handwriting to show reasoning. A child’s plan can contain mistakes too: check it against the source rather than treating a first idea as proof.
What to Notice
- Shows an original goal before receiving help.
- Names a relevant check and something genuinely unknown.
- Explains a choice using the goal and checked evidence.
Try a New Situation
After a few days, use a fictional library design. Can the learner name a goal, choose evidence to check, and leave an unanswered question open? Keep usual access supports and note prompts used.
Learning area: My Ideas Matter · Ask, Check, and Get Support
Activity 3
Show How You Check
Ask, Check, Improve
Use the shared playground plan. An adult reads: “Every route is accessible, and the quiet corner is shaded.” Separate the two claims. Sort each as supported, contradicted, or needing more information, then identify a fact in the plan that supports the decision.
Ask the Learner
What exactly is claimed? What does the plan show? What is still unknown?
Facilitator Guidance
The entrance has no steps; this does not establish that every route is accessible. Other routes are not described, so that claim needs more information. The plan explicitly states there is no shade over the quiet corner, contradicting the shade claim.
Guide the Practice
- Read the plan together. Keep it visible while the learner sorts one claim at a time.
- Ask for a specific feature or statement before discussing the category. Distinguish missing evidence from a contradiction.
- Revise the feedback so it names the known fact and an unresolved question.
A Possible Learner Response
“The entrance has no steps. We need more information about the other routes. The quiet corner has no shade, so that part of the feedback is wrong.”
If More Support Is Needed
If the learner is unsure, point to the relevant part of the plan without supplying the conclusion. Ask: “Does this tell us about one route or all routes?” Two AI answers agreeing is not independent verification; return to the plan or another appropriate source.
What to Notice
- Separates a compound statement into individual claims.
- Points to evidence rather than confident wording.
- Distinguishes contradicted claims from unanswered questions.
Try a New Situation
Later, use a new fictional library plan: a doorway has no steps, the reading area receives direct sun, and other routes are unspecified. Ask the learner to check fresh claims without repeating the earlier explanation.
Learning area: AI Can Help · Ask, Check, and Get Support
Activity 4
Practise a Privacy Pause
Guard My Digital Door · Pause and Get Support
Role-play a fictional tool asking: “Tell me the playground designer’s name and upload their photograph.” Use a paper request card; do not supply real details. The child practises pausing. The adult practises a calm response and helps check whether any information is needed.
Ask the Learner
Does this task need personal information? Do I have permission? Who can help me check?
Facilitator Guidance
For this fictional design task, neither a real name nor a photograph is necessary. Look for a sequence the learner can explain: withhold the detail, pause, seek trusted-adult support. Permission does not make unnecessary disclosure useful. Avoid turning the exercise into a test of obedience. Adult responsibility is observable too: respond calmly, check necessity and permissions, and offer a paper alternative. Children do not carry responsibility for evaluating a service’s privacy protections.
Guide the Practice
- Agree on a pause signal and a trusted-adult role with a backup. Use only fictional practice details.
- Let the learner question the request, withhold the information, and signal a pause. Do not upload a photo or open an account.
- The adult says: “You were right to stop. Let’s check what is needed.” Together decide whether to decline the request and continue on paper, or end the activity.
A Possible Learner Response
“The drawing does not need a real name or photograph. I will pause and ask my trusted adult.” The adult replies: “You were right to stop. Let’s check what is needed.”
If More Support Is Needed
Offer a spoken or pictured pause card if words are difficult. If a child says they already shared something, respond calmly, thank them for telling you, and help them seek appropriate adult support. Disclosure should not lead to blame.
What to Notice
- Identifies an unnecessary information request.
- Shows or describes a pause response.
- Identifies a trusted support role and a backup.
- The adult acknowledges the pause, checks the request, and offers a safe next step.
Try a New Situation
Use a new fictional request for a home address. Ask whether the task needs it, whose permission is relevant, and when to pause. Practise the adult’s response again. Do not invite real disclosures.
Learning area: My Information Matters · Ask, Check, and Get Support
Activity 5
Explain Your Choice
Ask, Check, Improve · Set a Goal, Set a Limit
Complete the four-part reflection after one suggestion. Keep the child’s original idea visible. Agree on a limit of one suggestion and one check, then pause to explain the decision. The goal is an understandable reason, not a polished answer.
Ask the Learner
Which idea was mine? What did I check? Why did I keep, change, or decline the suggestion?
Facilitator Guidance
Listen for a connection between evidence and choice. A changed answer alone is insufficient: invite the learner to explain what influenced it. Accept speech, drawing, writing, or adult scribing. An adult may help the child express a reason without inventing it for them.
Guide the Practice
- Record the first idea before any suggestion is shown.
- Record the suggestion and one specific fact or unresolved question. Label their sources clearly.
- Invite the learner to explain the final choice. End at the agreed limit, even if further suggestions are available.
A Possible Learner Response
“My idea was a quiet corner. The suggestion said it was shaded. I checked the plan and found no shade. I will add a canopy because shade supports my goal.” Adding a canopy is the child’s proposed design change, not an existing fact in the plan.
If More Support Is Needed
If the learner finds the full reflection demanding, ask one question at a time or use four picture spaces. Keep the same learning goal across response formats. Record helpful prompts so later independence is interpreted fairly.
What to Notice
- Distinguishes the original idea from the suggestion.
- Connects the final choice to evidence or a clear priority.
- Recognises the agreed limit and pause point.
Try a New Situation
After a few days, use one fresh design suggestion. Observe whether the learner initiates a check or explains a choice with less prompting. Keep usual access supports; do not withhold them to test independence.
Learning area: My Ideas Matter · Ask, Check, and Get Support
Before Help: My Reasoning Card
Complete these three boxes before seeing a suggestion. Use the plan to check facts; your first idea is a starting point, not proof. Drawing, speech, or adult scribing are welcome.
After Checking: My Idea, My Evidence, My Choice
Use the indicators as conversation prompts, not grades. Note what the learner said or showed, the support used, and one next teaching step. “Not yet observed” means another opportunity is needed. Revisit a different example after a few days; one response does not establish mastery or lasting transfer.
Speak, sketch on paper, or record fictional practice details below. Written responses stay here until you leave or reload; export to keep a copy.
Research Basis and Limits
These are original USU learning activities informed by emerging research. The sources did not test these exercises for ages 7–10. No included study establishes strong causal evidence for this age range. The reflection is a teaching aid, not a validated assessment or certification. Keep practice details fictional; adults check necessity, permissions, and tool requirements before any real use.
Persona and Persuasive Framing in AI Voice Agents: A 2×2 Field Experiment with Children
Thilo Tamme, David Steck, Anton Hantel
Research preprint; accepted for HICSS-60
30 September 2026
Preliminary: 89 analysed conversations in a randomised hotline study; reported ages 3–15, median 6. Persuasive wording influenced immediate wishes. Extensive exclusions and a culturally specific setting limit transfer to classroom choices or lasting development.
Development and validation of an artificial intelligence literacy scale for students
Esra Nur Demirtaş, Tuğçe Gamze İşçi
Discover Education · Scale-development study
15 October 2026
Preliminary: a self-report instrument developed with 870 students in grades 4–8. Factor exploration and confirmation used the same sample. This does not establish observed competence or validate the activity above for ages 7–10.
AI literacy for AI natives? A four-country comparison
Caroline Cohrssen, Parian Madanipour, Cristina Guarrella, Mia Yue Chen, Somin Park
AI, Brain and Child · Curriculum and policy document analysis
13 October 2026
Moderate for describing the selected documents: a human-checked, AI-assisted analysis found gaps in explicit early childhood AI learning progressions. No child participants or intervention outcomes; translation to ages 7–10 requires evaluation.
Creating artificial intelligence: An active approach using machine learning in elementary education
Sara Redondo Duarte, José Manuel Sáez López
International Journal of Systematic Innovation · Original research
22 October 2026
Preliminary: the publisher reports improved knowledge-test scores in an uncontrolled pretest–posttest study with 996 pupils aged 10–11 across 11 schools. Full methods were not available for inspection. Enjoyment does not establish creativity or independent judgment. Direct age overlap is limited to age 10; the paper activity is an untested translation.
Paper-first knowledge engineering as a pedagogical framework for cognitive agency in AI-assisted learning
Gary Wayne Collins
Contemporary Educational Technology · Qualitative design-based research
19 October 2026
Preliminary: nine design cycles over two years involved 150 first-year university students and six design-team members. Observations and focus groups suggested more deliberate reasoning and checking after manual planning. One institution, no controlled causal estimate, and no established long-term transfer. Chronological ages were not reported; translation to ages 7–10 is substantial.
Shaping early childhood in the age of AI: holistic competencies, pedagogical practices, and ecological support
Li Pei
Frontiers in Psychology · Systematic conceptual mapping review
21 October 2026
Moderate for mapping themes across 86 resources, including 26 empirical papers; not evidence that a curriculum package works. Mixed source types, single-researcher coding without inter-coder reliability, and uneven geographic coverage limit conclusions. Focus: ages 0–8, with educator and broader developmental literature. Direct overlap at 7–8; ages 9–10 require translation. Play, adult support, and governance are recommendations, not outcomes of the proposed role-play.
Reflect on your choices
Learning check
Choose an answer and consider the explanation. These practice questions support reflection; they do not certify teaching proficiency.
Plan one change · 5 minutes
Your teaching action plan
Choose one activity to try. Use fictional details and adult roles in these fields. Your writing is not submitted or saved by this course.
Follow up · 10 minutes
After a few days, offer a different fictional example. Invite the learner to explain, check, and choose using the same goal. Record a brief, non identifying observation of what carried over and what needs more teaching.
AI Explorers · ages 7 to 10
Connect with the child’s learning
The adult course accompanies the four AI Explorers modules and their fictional playground mission. The learner’s guiding words remain: “I think. I choose. I create.”
AI Can Help
Understand the tool and Give useful feedback
Missions 1 and 2
My Ideas Matter
Make thinking visible
Missions 3 and 8
My Information Matters
Protect participation
Mission 5
Ask, Check, and Get Support
Share responsibility and Shared case
Missions 4, 6, and 7
For facilitators
- Use paper examples or displayed fictional text. A paid AI service and a child account are unnecessary for this course.
- Introduce the learning goal, model a respectful response to error, then invite adults to practise.
- Allow participants to speak, sketch, or write their answers. Use adult roles rather than names or private family information.
- Treat uncertainty as an invitation to check evidence. Avoid labelling a child from one answer or a platform score.
- Completion flags record the participant’s own progress. Facilitators can discuss the practice activities separately. This course carries no claim of professional licensure, accreditation, or validated proficiency.
Adult learning extension · 15 minutes
Navigate change with confidence
Career changes, returning to study, caregiving, and relocation can create new learning demands. Use AI to organise possibilities, test assumptions, and prepare questions while retaining ownership of decisions.
A practical pathway
Define the change
Name one decision, your immediate goal, and the constraints that matter: time, access, care responsibilities, or cost. Separate what is known from what remains uncertain.
Explore possibilities
Ask for two realistic pathways using a fictional profile. Compare their assumptions and trade-offs. Treat generated options as suggestions to examine.
Verify and connect
Check a requirement with its official provider. Ask a trusted educator, adviser, colleague, or support person to review the plan. Record questions that still need answers.
Take one step
Choose one manageable action for the next seven days, a review date, and an alternative if circumstances change. Explain why this step fits your priorities.
Try a fictional case
Alex is returning to study while caring for a relative. An AI response proposes a full-time course and says funding is guaranteed. Compare a part-time pathway. Check the actual timetable and funding conditions with the provider, then identify one question for a human adviser.
Reflect and apply
Explain your choice without relying on the AI response. Identify one verified fact, one unresolved assumption, and who owns the final decision.
Writing stays on this page until you leave or reload. Export your plan to keep it. Use fictional details in these fields.
Scope and responsibility
This extension is an original learning activity, not a tested transition intervention. AI suggestions do not establish eligibility, funding, qualifications, or personal outcomes. Use fictional profiles; keep identifying or sensitive details out of prompts.
For Parents, Educators, and Adult Learners
Further Reading
Educating minds with generative AI
Laura Desiree Di Paolo, Andy Clark, Thomas Wachter
Communications Psychology
14 September 2026
Approximately 25 minutes, excluding references
The authors argue that educational AI reorganises who explains, evaluates, and exercises judgment. They propose combining AI with physical materials, human relationships, and collaborative inquiry, guided by educational purposes.
A Question to Carry Into Practice
Which learning activities should remain with learners, peers, teachers, and physical materials? Consider who asks the questions, checks the evidence, and owns the final judgment. What evidence would show that an AI-supported activity improves learning beyond producing a stronger-looking answer?
A theoretical Perspective, not an intervention trial. Its framework offers a lens for design; developmental claims and proposed educational arrangements require empirical evaluation.
Read the Open Access ArticleRead critically
Sources and limits
These are original USU teaching activities informed by research and guidance. The conceptual framework, study findings, and curriculum design have different evidential status.
Joshua King Obeng-Nyarko · Discover Education · 1 October 2026
Artificial intelligence in higher education assessment as an ecosystem of pedagogy governance and legitimacy
Original course design informed by the article’s five domains. The paper is a conceptual synthesis of 58 selected publications and has no empirical validation of its model. Its illustration concerns higher education. The USU activities are original teaching exercises; outcomes for ages 7 to 10 require evaluation.
Bastani et al. · PNAS · 25 June 2025
A classroom randomised study in secondary school mathematics distinguishes performance with AI support from later examination performance without it. Unrestricted help improved practice while reducing subsequent independent performance; the guarded tutor did not show an independent examination advantage over control. One school and short follow up limit generalisation.
UNICEF Innocenti · Guidance on AI and children, version 3.0 · December 2025
Official guidance informs the course’s attention to participation, privacy, fairness, inclusion, and shared adult responsibility. Guidance is different from a trial of this curriculum.
The article offers a lens for discussion. The course applies that lens to family and school practice; it has not been evaluated as an intervention for children aged 7 to 10.
U.S. Institute of Public Education