See what children learn, why each lab is built this way, and the sources behind the design.
019 principles, 15 checked sources
02Every source shows its check status
03No efficacy claim
Figure 1Principles and the mechanics they shaped
Choose a principle to see what it shaped
Principles
Mechanics
Mechanics
M1 Peek Hatch locked until guess
M3 Tweak one thing, compare
M2 Seven-step Mechna Loop
M3 Tweak one thing, compare
M4 Three-level hints on request
M5 A case never seen
M1 Peek Hatch locked until guess
M6 Discovery Card with evidence
M7 Picture-depth Peek Hatch
M8 States, not feelings
M9 Plate: who made it, why
M10 A human guide, not a robot
M11 Voice, captions, replay
M12 No timers, lives or losing
P1
Guess first, then look
For parents:
When children commit to a guess before they see how something works, the answer tends to stick better, especially when it surprises them.
The mechanic it shaped:
The Peek Hatch, which shows how a machine decides, stays locked until the child picks a guess. “Can’t tell yet” counts as a real answer.
Honest limit
C01 studied university students. C03 is teacher-led programming for ages 11–14. Whether the benefit holds for 8–10-year-olds in a self-paced app is our hypothesis, not a finding.
C01VerifiedBrod, G., Hasselhorn, M., & Bunge, S. A. (2018). When generating a prediction boosts learning: The element of surprise. Learning and Instruction, 55, 22–31.
C02Partial: abstract checkedBrod, G. (2021). Predicting as a learning strategy. Psychonomic Bulletin & Review, 28(6), 1839–1847.
C03Partial: abstract checkedSentance, S., Waite, J., & Kallia, M. (2019). Teaching computer programming with PRIMM: A sociocultural perspective. Computer Science Education, 29(2–3), 136–176.
P2
One familiar routine in every lab
For parents:
Science teaching often follows a five-stage cycle (engage, explore, explain, elaborate, evaluate), so every Mechna lab follows the same routine that a child learns once and then uses in every lab.
The mechanic it shaped:
The seven-step Mechna Loop runs in every lab: Spot, Try, Guess, Peek, Tweak, Prove, Keep.
Honest limit
The 5E model was designed for teacher-led science classes. Mapping it onto a self-paced lab is our design decision.
C04VerifiedBybee, R. W., Taylor, J. A., Gardner, A., Van Scotter, P., Carlson Powell, J., Westbrook, A., & Landes, N. (2006). The BSCS 5E instructional model: Origins and effectiveness. BSCS.
P3
A fair test changes one thing
For parents:
To find out what really causes a result, you change one thing at a time. Children aged 7 to 10 have learned this fair-test habit when it was taught clearly.
The mechanic it shaped:
The Tweak step lets a child change one thing, guess what will happen, then compare before and after.
Honest limit
Both studies used physical-science tasks such as ramps, not AI machines.
C05Partial: abstract checkedChen, Z., & Klahr, D. (1999). All other things being equal: Acquisition and transfer of the control of variables strategy. Child Development, 70(5), 1098–1120.
C06VerifiedKlahr, D., & Nigam, M. (2004). The equivalence of learning paths in early science instruction: Effects of direct instruction and discovery learning. Psychological Science, 15(10), 661–667.
P4
Help is there whenever a child asks
For parents:
Beginners learn well from a clear worked example, so step-by-step help is always one tap or click away and never counts against a child.
The mechanic it shaped:
Hints come in three levels, from a small nudge to a full worked example, and a child asks for them.
Honest limit
The sources support worked examples and explicit teaching. Neither tests hints that a child asks for inside an app. C06 also found that direct teaching beat discovery for this skill, which a qualified curriculum reviewer will weigh against a loop that starts with exploring.
C07Verified · Recommendation 2Pashler, H., Bain, P. M., Bottge, B. A., Graesser, A., Koedinger, K., McDaniel, M., & Metcalfe, J. (2007). Organizing instruction and study to improve student learning (NCER 2007-2004). Institute of Education Sciences, U.S. Department of Education.
C06VerifiedKlahr, D., & Nigam, M. (2004). The equivalence of learning paths in early science instruction: Effects of direct instruction and discovery learning. Psychological Science, 15(10), 661–667.
P5
Prove it on something new
For parents:
Understanding shows when a child can use an idea on a case they’ve never seen, not when they repeat an answer.
The mechanic it shaped:
The Prove step ends each lab with a fresh case the machine has never met. The child guesses first, then checks.
Honest limit
C08 shows that transfer depends heavily on how different the new case is. A fresh case in the same lab is a near test, not proof of broad transfer.
C05Partial: abstract checkedChen, Z., & Klahr, D. (1999). All other things being equal: Acquisition and transfer of the control of variables strategy. Child Development, 70(5), 1098–1120.
C08Partial: abstract checkedBarnett, S. M., & Ceci, S. J. (2002). When and where do we apply what we learn? A taxonomy for far transfer. Psychological Bulletin, 128(4), 612–637.
P6
Explain it in your own words, with evidence
For parents:
Children learn more deeply when they build an explanation themselves than when they only watch or click.
The mechanic it shaped:
At the end of a lab, the child builds a Discovery Card: a claim put together from word tiles and linked to the evidence they used.
Honest limit
A claim built from tiles is a supported explanation, not a free one. It sits toward the “constructive” end of C09’s scale, but we haven’t measured it.
C09VerifiedChi, M. T. H., & Wylie, R. (2014). The ICAP framework: Linking cognitive engagement to active learning outcomes. Educational Psychologist, 49(4), 219–243.
C07Verified · Recommendation 7Pashler, H., Bain, P. M., Bottge, B. A., Graesser, A., Koedinger, K., McDaniel, M., & Metcalfe, J. (2007). Organizing instruction and study to improve student learning (NCER 2007-2004). Institute of Education Sciences, U.S. Department of Education.
P7
From pictures to code
For parents:
Children understand abstract ideas better when they start from concrete pictures and objects that are clearly linked to the abstract version.
The mechanic it shaped:
For ages 8–10, the Peek Hatch shows how a machine decides with picture cards and objects.
Honest limit
C07 is about linking concrete and abstract forms within teaching. Spreading depths across age groups is our extension of it. C10 is a working draft.
C07Verified · Recommendation 4Pashler, H., Bain, P. M., Bottge, B. A., Graesser, A., Koedinger, K., McDaniel, M., & Metcalfe, J. (2007). Organizing instruction and study to improve student learning (NCER 2007-2004). Institute of Education Sciences, U.S. Department of Education.
C10VerifiedAI4K12 Initiative (AAAI & CSTA). (2020). Draft Big Idea 3: Progression chart (v0.1, released November 19, 2020; “subject to change based on public feedback”).
P8
Machines are made by people, and they don’t have feelings
For parents:
Children often treat talking devices as clever beings. Good AI education shows that machines are built by people, for a purpose, and don’t feel anything.
The mechanic it shaped:
Every machine shows states (ready, running, stuck), never feelings, and has no face.
Every machine has a plate saying who made it and why.
The guide is a human character, a senior apprentice, not a robot friend.
Honest limit
C12 is a small exploratory study (ages 3–10). The frameworks say what children should understand, but don’t test our design.
C11VerifiedTouretzky, D., Gardner-McCune, C., Martin, F., & Seehorn, D. (2019). Envisioning AI for K-12: What should every child know about AI? Proceedings of the AAAI Conference on Artificial Intelligence, 33(01), 9795–9799.
C12Partial: abstract checkedDruga, S., Williams, R., Breazeal, C., & Resnick, M. (2017). “Hey Google is it OK if I eat you?” Initial explorations in child-agent interaction. In Proceedings of the 2017 Conference on Interaction Design and Children (pp. 595–600). ACM.
C13VerifiedMiao, F., Shiohira, K., & Lao, N. (2024). AI competency framework for students. UNESCO.
P9
Many ways in, and no clock
For parents:
Children learn well from pictures with spoken explanation, and every child needs more than one way in: to hear it, see it and answer without racing a clock.
The mechanic it shaped:
Voice is on by default, captions are always shown, and any line can be replayed.
Children answer by tapping or clicking pictures and word tiles, with no typing.
There are no timers, lives or losing.
Honest limit
C14 and C15 are design and accessibility frameworks, not learning trials. Having narration and captions on together is a choice we made for access, and a qualified curriculum reviewer will check it.
C07Verified · Recommendation 3Pashler, H., Bain, P. M., Bottge, B. A., Graesser, A., Koedinger, K., McDaniel, M., & Metcalfe, J. (2007). Organizing instruction and study to improve student learning (NCER 2007-2004). Institute of Education Sciences, U.S. Department of Education.
C14VerifiedCAST. (2024). Universal Design for Learning Guidelines version 3.0.
C15VerifiedW3C. (2024). Web Content Accessibility Guidelines (WCAG) 2.2 (W3C Recommendation; this version dated December 12, 2024). A. Campbell, C. Adams, R. Bradley Montgomery, M. Cooper, & A. Kirkpatrick (Eds.).
Honest limit: C01 studied university students. C03 is teacher-led programming for ages 11–14. Whether the benefit holds for 8–10-year-olds in a self-paced app is our hypothesis, not a finding.
Honest limit: The sources support worked examples and explicit teaching. Neither tests hints that a child asks for inside an app. C06 also found that direct teaching beat discovery for this skill, which a qualified curriculum reviewer will weigh against a loop that starts with exploring.
P5 Prove it on something new
Shaped:A case never seenPeek Hatch locked until guess
Honest limit: C08 shows that transfer depends heavily on how different the new case is. A fresh case in the same lab is a near test, not proof of broad transfer.
Honest limit: A claim built from tiles is a supported explanation, not a free one. It sits toward the “constructive” end of C09’s scale, but we haven’t measured it.
Honest limit: C07 is about linking concrete and abstract forms within teaching. Spreading depths across age groups is our extension of it. C10 is a working draft.
P8 Machines are made by people, and they don’t have feelings
Shaped:States, not feelingsPlate: who made it, whyA human guide, not a robot
Honest limit: C14 and C15 are design and accessibility frameworks, not learning trials. Having narration and captions on together is a choice we made for access, and a qualified curriculum reviewer will check it.
/// ENTRY 002 _ WHAT WE CLAIM
What we claim, and what we don’t.
Mechna Academy is a learning app where children test how AI machines work. They never talk to a live AI.
This page sets out the learning science behind Mechna. Sources explain our design choices; they do not establish that Mechna improves learning.
Research sources explain Mechna Academy’s design choices; they do not establish that Mechna improves learning.
We make no efficacy claim or percentage about Mechna.
/// ENTRY 003 _ PRINCIPLE TO MECHANIC
Why it’s built the way it is.
Nine principles, each with the mechanic it shaped, its sources and its honest limit. Sources marked “abstract checked” were checked against their abstract or a registry record only.
P1
P1. Guess first, then look
What the research shows
Making a prediction before seeing the answer can support learning, especially when the result is surprising (C01).
Who it was studied with
C01 studied university students. C03 examined teacher-led programming with ages 11–14. Whether this benefit holds for 8–10-year-olds in a self-paced app is a hypothesis, not a finding.
For parents:
When children commit to a guess before they see how something works, the answer tends to stick better, especially when it surprises them.
The mechanic it shaped:
The Peek Hatch, which shows how a machine decides, stays locked until the child picks a guess. “Can’t tell yet” counts as a real answer.
Brod, G., Hasselhorn, M., & Bunge, S. A. (2018). When generating a prediction boosts learning: The element of surprise. Learning and Instruction, 55, 22–31. doi.org/10.1016/j.learninstruc.2018.01.013
Sentance, S., Waite, J., & Kallia, M. (2019). Teaching computer programming with PRIMM: A sociocultural perspective. Computer Science Education, 29(2–3), 136–176. doi.org/10.1080/08993408.2019.1608781
P2
P2. One familiar routine in every lab
What the research shows
The BSCS 5E instructional model sets out a five-stage cycle for science teaching (C04).
Who it was studied with
C04 describes a model designed for teacher-led science classes. Mapping it onto a self-paced lab is our design decision, not a finding about this app.
For parents:
Science teaching often follows a five-stage cycle (engage, explore, explain, elaborate, evaluate), so every Mechna lab follows the same routine that a child learns once and then uses in every lab.
The mechanic it shaped:
The seven-step Mechna Loop runs in every lab: Spot, Try, Guess, Peek, Tweak, Prove, Keep.
Bybee, R. W., Taylor, J. A., Gardner, A., Van Scotter, P., Carlson Powell, J., Westbrook, A., & Landes, N. (2006). The BSCS 5E instructional model: Origins and effectiveness. BSCS. bscs.org
P3
P3. A fair test changes one thing
What the research shows
Children aged 7 to 10 have learned to change one thing at a time in fair tests when taught clearly (C05, C06).
Who it was studied with
The cited work concerns children aged 7 to 10 learning with physical-science tasks such as ramps, not machines in this app.
For parents:
To find out what really causes a result, you change one thing at a time. Children aged 7 to 10 have learned this fair-test habit when it was taught clearly.
The mechanic it shaped:
The Tweak step lets a child change one thing, guess what will happen, then compare before and after.
Chen, Z., & Klahr, D. (1999). All other things being equal: Acquisition and transfer of the control of variables strategy. Child Development, 70(5), 1098–1120. doi.org/10.1111/1467-8624.00081
Klahr, D., & Nigam, M. (2004). The equivalence of learning paths in early science instruction: Effects of direct instruction and discovery learning. Psychological Science, 15(10), 661–667. doi.org/10.1111/j.0956-7976.2004.00737.x
How we’re checking it
“Does a change the child makes connect to the result they see?” Research methods
P4
P4. Help is there whenever a child asks
What the research shows
Worked examples and explicit teaching can support beginners learning a new skill (C07, C06).
Who it was studied with
C07 and C06 address instruction, but neither tests hints that a child asks for inside an app. C06 found that direct teaching beat discovery for this skill; a qualified curriculum reviewer will weigh that against a loop that starts with exploring.
For parents:
Beginners learn well from a clear worked example, so step-by-step help is always one tap or click away and never counts against a child.
The mechanic it shaped:
Hints come in three levels, from a small nudge to a full worked example, and a child asks for them.
Pashler, H., Bain, P. M., Bottge, B. A., Graesser, A., Koedinger, K., McDaniel, M., & Metcalfe, J. (2007). Organizing instruction and study to improve student learning (NCER 2007-2004). Institute of Education Sciences, U.S. Department of Education. ies.ed.gov
Klahr, D., & Nigam, M. (2004). The equivalence of learning paths in early science instruction: Effects of direct instruction and discovery learning. Psychological Science, 15(10), 661–667. doi.org/10.1111/j.0956-7976.2004.00737.x
P5
P5. Prove it on something new
What the research shows
Transfer to a new case depends heavily on how different that case is from what was learned (C08).
Who it was studied with
C05 used physical-science tasks with children aged 7 to 10. C08 describes transfer across cases; a fresh case in the same lab is a near test, not proof of broad transfer.
For parents:
Understanding shows when a child can use an idea on a case they’ve never seen, not when they repeat an answer.
The mechanic it shaped:
The Prove step ends each lab with a fresh case the machine has never met. The child guesses first, then checks.
Chen, Z., & Klahr, D. (1999). All other things being equal: Acquisition and transfer of the control of variables strategy. Child Development, 70(5), 1098–1120. doi.org/10.1111/1467-8624.00081
Barnett, S. M., & Ceci, S. J. (2002). When and where do we apply what we learn? A taxonomy for far transfer. Psychological Bulletin, 128(4), 612–637. doi.org/10.1037/0033-2909.128.4.612
How we’re checking it
“Can the child reason about a case they haven’t seen?” Research methods
P6
P6. Explain it in your own words, with evidence
What the research shows
Building an explanation can support deeper learning than only watching or clicking (C09).
Who it was studied with
C09 describes a scale of cognitive engagement, not a test of a claim built from tiles. That supported explanation sits toward the constructive end of the scale, but we haven’t measured it.
For parents:
Children learn more deeply when they build an explanation themselves than when they only watch or click.
The mechanic it shaped:
At the end of a lab, the child builds a Discovery Card: a claim put together from word tiles and linked to the evidence they used.
Chi, M. T. H., & Wylie, R. (2014). The ICAP framework: Linking cognitive engagement to active learning outcomes. Educational Psychologist, 49(4), 219–243. doi.org/10.1080/00461520.2014.965823
Pashler, H., Bain, P. M., Bottge, B. A., Graesser, A., Koedinger, K., McDaniel, M., & Metcalfe, J. (2007). Organizing instruction and study to improve student learning (NCER 2007-2004). Institute of Education Sciences, U.S. Department of Education. ies.ed.gov
P7
P7. From pictures to code
What the research shows
Teaching that links concrete pictures and objects to abstract forms can help students understand abstract ideas (C07).
Who it was studied with
C07 addresses linked concrete and abstract forms within teaching. C10 is a working draft. Spreading depths across age groups is our extension of that guidance, not a finding.
For parents:
Children understand abstract ideas better when they start from concrete pictures and objects that are clearly linked to the abstract version.
The mechanic it shaped:
For ages 8–10, the Peek Hatch shows how a machine decides with picture cards and objects.
Pashler, H., Bain, P. M., Bottge, B. A., Graesser, A., Koedinger, K., McDaniel, M., & Metcalfe, J. (2007). Organizing instruction and study to improve student learning (NCER 2007-2004). Institute of Education Sciences, U.S. Department of Education. ies.ed.gov
AI4K12 Initiative (AAAI & CSTA). (2020). Draft Big Idea 3: Progression chart (v0.1, released November 19, 2020; “subject to change based on public feedback”). ai4k12.org
P8
P8. Machines are made by people, and they don’t have feelings
What the research shows
Education frameworks call for children to understand that machines are made by people for a purpose (C11, C13).
Who it was studied with
C12 was a small exploratory study with children aged 3–10. C11 and C13 are frameworks about what children should understand; they do not test our design.
For parents:
Children often treat talking devices as clever beings. Good AI education shows that machines are built by people, for a purpose, and don’t feel anything.
The mechanic it shaped:
Every machine shows states (ready, running, stuck), never feelings, and has no face.
Every machine has a plate saying who made it and why.
The guide is a human character, a senior apprentice, not a robot friend.
Touretzky, D., Gardner-McCune, C., Martin, F., & Seehorn, D. (2019). Envisioning AI for K-12: What should every child know about AI? Proceedings of the AAAI Conference on Artificial Intelligence, 33(01), 9795–9799. doi.org/10.1609/aaai.v33i01.33019795
Druga, S., Williams, R., Breazeal, C., & Resnick, M. (2017). “Hey Google is it OK if I eat you?” Initial explorations in child-agent interaction. In Proceedings of the 2017 Conference on Interaction Design and Children (pp. 595–600). ACM. doi.org/10.1145/3078072.3084330
Miao, F., Shiohira, K., & Lao, N. (2024). AI competency framework for students. UNESCO. unesco.org
P9
P9. Many ways in, and no clock
What the research shows
Instructional guidance supports pictures with spoken explanation, while accessibility frameworks call for more than one way to access learning (C07, C14, C15).
Who it was studied with
C07 is instructional guidance; C14 and C15 are design and accessibility frameworks, not learning trials. Narration and captions together are our choice for access, which a qualified curriculum reviewer will check.
For parents:
Children learn well from pictures with spoken explanation, and every child needs more than one way in: to hear it, see it and answer without racing a clock.
The mechanic it shaped:
Voice is on by default, captions are always shown, and any line can be replayed.
Children answer by tapping or clicking pictures and word tiles, with no typing.
Pashler, H., Bain, P. M., Bottge, B. A., Graesser, A., Koedinger, K., McDaniel, M., & Metcalfe, J. (2007). Organizing instruction and study to improve student learning (NCER 2007-2004). Institute of Education Sciences, U.S. Department of Education. ies.ed.gov
W3C. (2024). Web Content Accessibility Guidelines (WCAG) 2.2 (W3C Recommendation; this version dated December 12, 2024). A. Campbell, C. Adams, R. Bradley Montgomery, M. Cooper, & A. Kirkpatrick (Eds.). w3.org/TR/WCAG22
How we’re checking it
“Where do reading, navigation, touch or saving get in the way? These are recorded apart from understanding.” Research methods
/// ENTRY 005 _ RESEARCH METHODS
What research can learn from.
Parent feedback and observed sessions answer different questions. Observation requires separate consent.
At home: parent feedback
Children aged 8 to 10 use Mechna at home on an iPad or Mac. Progress stays on the device.
What we want to learnWhere it comes from
How families describe their experience
Parent feedback
Whether children choose to come back on their own
Parent report
Where children got stuck or lost interest
Parent feedback
Whether the app runs reliably
Device reports and parent feedback
Device reports do not measure a child’s learning.
Parent report is what a parent noticed. We don’t read it as a measure of learning.
What we won’t collect at home
Data from inside Mechna: it doesn’t connect to the internet
Test scores, grades or mastery levels
Recordings, voice or photos of children
A child’s name or birth date
Optional observed sessions
An observed session requires separate written consent and remains optional.
An observed session uses a written protocol, a facilitator and a parent present. The child chooses whether to take part and can stop at any time.
Observation notes do not appear on this site.
What an observed session looks for
Goal
Can the child say what they’re trying to find out, without an adult explaining it again?
Cause
Does a change the child makes connect to the result they see?
A fresh case
Can the child reason about a case they haven’t seen?
Choosing to go on
Offered an even choice to continue or stop, does the child choose another experiment? This is recorded separately from learning.
Access and controls
Where do reading, navigation, touch or saving get in the way? These are recorded apart from understanding.
/// ENTRY 006 _ CONSENT
Consent comes first.
Research participation requires a parent or guardian’s agreement, and an observed session requires separate written consent.
01
At home
Research participation requires a parent or guardian’s plain-language agreement about the information collected.
Children receive a plain-language explanation and can stop at any time.
Participants can leave at any time and ask to delete their email and answers.
02
Optional observed sessions
An observed session requires separate written consent and remains optional.
Your child is asked at the start whether they’d like to take part, and can stop at any point. A parent is there throughout.
We never publish a child’s name, image or work without separate, specific permission. Research participation does not give that permission.
Research reporting describes methods and limits in plain language.
What we set out to learn
What families and observed sessions showed us about the design, in words, including what didn’t work
What we changed as a result
What we still don’t know
We do not publish child-level data, scores, percentages, quotes or images of children.
/// ENTRY 008 _ LIMITS
Claims we avoid.
That Mechna improves learning without evidence
That one child’s or one family’s experience stands for all children
That children “mastered” anything, or earned a grade or placement from research
That at-home observations are a controlled study
Anything about a named, pictured or identifiable child
A small group cannot show that learning lasts or that every mission works for every child.
/// ENTRY 009 _ REFERENCES
References.
All 15 sources were checked on September 29, 2026. Verified: checked against the full text or official page. Abstract checked: only the abstract or a registry record was available.
Verified
Abstract checked
10 verified · 5 abstract checked
C01
Verified
Brod, G., Hasselhorn, M., & Bunge, S. A. (2018). When generating a prediction boosts learning: The element of surprise. Learning and Instruction, 55, 22–31. doi.org/10.1016/j.learninstruc.2018.01.013
Sentance, S., Waite, J., & Kallia, M. (2019). Teaching computer programming with PRIMM: A sociocultural perspective. Computer Science Education, 29(2–3), 136–176. doi.org/10.1080/08993408.2019.1608781
C04
Verified
Bybee, R. W., Taylor, J. A., Gardner, A., Van Scotter, P., Carlson Powell, J., Westbrook, A., & Landes, N. (2006). The BSCS 5E instructional model: Origins and effectiveness. BSCS. bscs.org
C05
Partial: abstract checked
Chen, Z., & Klahr, D. (1999). All other things being equal: Acquisition and transfer of the control of variables strategy. Child Development, 70(5), 1098–1120. doi.org/10.1111/1467-8624.00081
C06
Verified
Klahr, D., & Nigam, M. (2004). The equivalence of learning paths in early science instruction: Effects of direct instruction and discovery learning. Psychological Science, 15(10), 661–667. doi.org/10.1111/j.0956-7976.2004.00737.x
C07
Verified
Pashler, H., Bain, P. M., Bottge, B. A., Graesser, A., Koedinger, K., McDaniel, M., & Metcalfe, J. (2007). Organizing instruction and study to improve student learning (NCER 2007-2004). Institute of Education Sciences, U.S. Department of Education. ies.ed.gov
C08
Partial: abstract checked
Barnett, S. M., & Ceci, S. J. (2002). When and where do we apply what we learn? A taxonomy for far transfer. Psychological Bulletin, 128(4), 612–637. doi.org/10.1037/0033-2909.128.4.612
C09
Verified
Chi, M. T. H., & Wylie, R. (2014). The ICAP framework: Linking cognitive engagement to active learning outcomes. Educational Psychologist, 49(4), 219–243. doi.org/10.1080/00461520.2014.965823
C10
Verified
AI4K12 Initiative (AAAI & CSTA). (2020). Draft Big Idea 3: Progression chart (v0.1, released November 19, 2020; “subject to change based on public feedback”). ai4k12.org
C11
Verified
Touretzky, D., Gardner-McCune, C., Martin, F., & Seehorn, D. (2019). Envisioning AI for K-12: What should every child know about AI? Proceedings of the AAAI Conference on Artificial Intelligence, 33(01), 9795–9799. doi.org/10.1609/aaai.v33i01.33019795
C12
Partial: abstract checked
Druga, S., Williams, R., Breazeal, C., & Resnick, M. (2017). “Hey Google is it OK if I eat you?” Initial explorations in child-agent interaction. In Proceedings of the 2017 Conference on Interaction Design and Children (pp. 595–600). ACM. doi.org/10.1145/3078072.3084330
C13
Verified
Miao, F., Shiohira, K., & Lao, N. (2024). AI competency framework for students. UNESCO. unesco.org
C14
Verified
CAST. (2024). Universal Design for Learning Guidelines version 3.0.udlguidelines.cast.org
C15
Verified
W3C. (2024). Web Content Accessibility Guidelines (WCAG) 2.2 (W3C Recommendation; this version dated December 12, 2024). A. Campbell, C. Adams, R. Bradley Montgomery, M. Cooper, & A. Kirkpatrick (Eds.). w3.org/TR/WCAG22
/// ANSWERS
Questions about the evidence.
Q01Is Mechna research-backed?
This page sets out the learning science Mechna’s design draws on: nine principles, each with its sources and the part of the game it shaped. Those sources explain our design choices; they don’t show that Mechna itself works. This page makes no efficacy claim. The design and its outcomes are separate questions.
Q02How does Mechna approach research?
Parent feedback and optional observed sessions offer different views of how children use Mechna. Observed sessions require separate consent. Neither is proof of learning gains.
Q03How does research consent work?
Research participation requires a parent or guardian’s agreement. Children receive a plain-language explanation and can stop at any time.
Q04Why does Mechna ask children to guess before they look?
When children commit to a guess before they see how something works, the answer tends to stick better, especially when it surprises them. In Mechna, the Peek Hatch that shows how a machine decides stays locked until the child picks a guess, and “can’t tell yet” counts as a real answer. Whether the benefit holds for 8–10-year-olds in a self-paced app is our hypothesis, not a finding.