Judgement first. Then AI.
Most AI literacy
starts with a screen.
We start with the hand.
Hands-on learning for sound judgement about AI before children begin using AI tools.
What the lens reveals The filter tracks shape. Other features remain outside this decision.
01
What is Sense & Think?
Hands-on AI literacy · ages 6–9
Sense & Think is a hands-on learning system that helps children aged 6–9 develop the judgement they need to use AI safely, before they begin using AI tools.
It combines physical activities, a teaching method and a clear process for educators. The screen is not the centre of learning.
What one lesson looks like
One prepared activity. Ten to fifteen minutes of focused work.
The teacher places one prepared activity on the table. Working alone or in a small group, a child spends 10–15 minutes sorting, testing and revising a decision. The material provides feedback, followed by a short reflection that transfers the principle to a new situation.
What children learn
Children learn that AI finds patterns in examples, can fail on a new case, and that claims need evidence. They learn when “I don’t know yet” is the most accurate answer, why not every piece of data should be shared, and why people must remain responsible for important decisions.
02
How a system finds patterns
How a systemfinds patternsin examples.
The system accepted three pieces. The examples alone do not reveal which rule it uses. Choose a new piece that distinguishes between the two rules.
Examples we have
All three were accepted.
All are circular and have lines. Their colour varies.
Two possible rules
Both still fit the examples.
Hypothesis ALINESshape does not matter
Hypothesis BCIRCLE + LINESshape matters
These data do not distinguish between them. We need an example for which the rules predict different outcomes.
Design a distinguishing test
Which new piece will reveal the difference between the rules?
The selected piece moves into the test position. Then compare the predictions made by both rules.
Choose a test that produces different outcomes for the two hypotheses.
Distinguishing test result
Only the square piece with lines reveals which of the two rules was used.
In this demonstration, we set the system to use the “lines” rule. The original examples could not reveal that rule unambiguously. A new test example was needed.
The new example exposed the weakness of the “CIRCLE + LINES” hypothesis.
What the child learns
The same examples can support several possible rules. That is why we need a new test.
The child compares hypotheses, chooses an example on which they diverge, and explains what the new result shows.
Simplified model: this activity demonstrates a hidden classification rule. It is not a literal simulation of how an AI model is trained.
03
Control of error
How the material enables self-correction
The child does not need to wait for an adult to point out the mistake. The material shows what does not fit.
Position, gaps, resistance and stability provide feedback. The child can revise the decision independently.
Place the piece. The material shows whether it fits clearly, does not fit, or remains ambiguous.
Concept visualisation · not a photograph of a finished prototype.
The frame is ready. Choose a piece and observe its position, gaps and stability.
04
When evidence is not enough
When evidence is missing, the answer may be I don’t know yet.
Sense & Think teaches children that a confident tone is not evidence. A safe decision may be yes, no — or a deliberate pause.
We now use the same borderline example for a different question: not which rule the system uses, but whether we have enough evidence to decide.
Not all information has the same value. Select the clues that are relevant to the claim, then decide.
1 · System claim
“The new square piece belongs in the group.”
The system did not reveal its rule. We only have the accepted examples and three possible clues.
2 · Select evidence
Which clues are actually relevant to the claim?
Select every relevant clue. Not every true piece of information is evidence.
No clue has been selected yet.
3 · Decide from the selection
Is the selected evidence enough to accept or reject the claim?
Select the relevant clues first. Then record your decision.
What else would you need to know to be more certain?
We could learn the rule used or test another borderline piece. A good check does not add certainty through tone — it adds relevant evidence.
What this trained
The child distinguishes between what they think and what the evidence supports. “I don’t know yet” is not failure; it is an accurate description of the situation.
Not every true piece of information is evidence for the question we are asking.
05
Privacy and purpose
A useful recommendation
does not require all your data.
Privacy is not a list of forbidden data. It depends on the purpose and the smallest amount of data needed to achieve it.
Choose a purpose, then select only the data needed for that purpose.
Who is asking?Reading app
Why is it needed?Recommend a suitable book
Your decisionChoose the smallest useful set of data.
Minimum data up to 4
Select only the data needed to recommend a book.
Privacy is not simply a list of forbidden data. It is a decision made in relation to a purpose.
06
System limits
When a person must decide
Some decisions should not be made by a system.
When context is missing, consequences may be serious, or a decision concerns a person, the process stops and an adult takes responsibility.
The system recommended a punishment based on a single incomplete behaviour record.
- the situation has not been explained
- the decision could harm the child
- a person is being judged, not a technical object
What must a person find out before making a decision?
Select every step that adds the missing context.
Choose a safe procedure, then check it.
A safe procedure adds the child’s voice, the full context and independent verification. A single incomplete record is not enough for a decision with serious consequences.
When evidence is insufficient or a decision may have serious consequences, the child stops and involves an adult.
07
Learning cycle
One task, five steps.
Every activity guides the child through the same cycle. We do not only correct the answer. We revise how the decision was made.
Observe the object first.
The child handles the piece and notices its shape, surface and differences without being given a verdict.
08
What we are testing
We test learning and transfer to a new situation.
We test what a child can do in a new situation.
We measure performance, transfer, independence, revision of judgement and teacher usability — not only enjoyment, satisfaction or reported understanding.
View eight areas of evaluation ＋
- 01Does the child understand that a system works with patterns?explanation in the child’s own words
- 02Can the child recognise when the original examples are insufficient?selection of a new test case
- 03Does the child say “I don’t know yet” when appropriate?task with incomplete information
- 04Can the child distinguish evidence from irrelevant information?new situation with different clues
- 05Does the child select only the data needed for the purpose?new privacy scenario
- 06Can the child recognise when a person must decide?high-stakes scenario or missing context
- 07Can a teacher run the activity without its author?teacher usability test
- 08Can the child use the principle in a new situation?transfer task with different content
Methodological foundations: Montessori control of error, children’s metacognition, screen-free AI literacy, evidence reasoning, human oversight and data minimisation.
Open the methodology overview ↗It explains the learning principles interactively. It is not a digital version of the final physical system.
First physical prototypes, methodological review and preparation of the pilot format.
Testing with children and teachers, revising materials and evaluating transfer of learning.
09
Pilot partnership
Preparing the first pilots
We are looking for the first pilot schools and expert partners.
For children aged 6–9 in Montessori and mainstream primary schools. The pilot will begin after methodological review of the first prototypes. We are involving interested partners now.
What the pilot involves ＋
- 1–2 classes or small groups
- children aged 6–9
- teachers do not need a technical AI background
- 2–3 shorter sessions
- Sense & Think provides the materials and teaching method
- teacher feedback and anonymised performance tasks
- early access and methodological support
- the opportunity to shape the final system
- a summary of pilot findings
- space for delivery and teacher cooperation
- constructive feedback
- compliance with child safeguarding and consent requirements
10
Partners and support
For partners and supporters
Support development of Sense & Think.
Support helps fund physical prototype development, methodological review, school pilots and evaluation of learning outcomes.
12
Contact