Pith. sign in

REVIEW 3 major objections 5 minor 41 references

A Structured Unplugged Approach for Foundational AI Literacy in Primary Education

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that a structured, screen-free curriculum anchored in mathematics improves foundational AI literacy in fifth-grade students.

desk verdict A well-structured unplugged AI curriculum whose effectiveness claims outrun its evidence. read the letter →

arxiv 2505.21398 v1 pith:2SU3YYPH submitted 2025-05-27 cs.AI cs.ET

Maria Cristina Carrisi , Mirko Marras , Sara Vergallo This is my paper
classification cs.AIcs.ET
keywords AIliteracyunpluggedlearningprimaryeducationclassificationdecisiontreesdatarepresentationK-12computingmathematicsintegration
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes a structured, screen-free learning path that introduces primary-school children to AI through mathematical ideas already present in their curriculum: classification, sets, frequencies, and data representation. It claims that after four two-hour modules, fifth-grade students improved in using AI terminology, describing features, reasoning logically about classification, and evaluating AI's errors and limits, while staying engaged. The point of the claim is that AI literacy does not require screens or programming exercises; it can be built from concrete classification and decision-tree activities that teachers can replicate. This matters because children meet AI daily through assistants, recommendations, and games, and without conceptual foundations they tend to treat these systems as intelligent and infallible.

What carries the argument

The load-bearing mechanism is a four-module, screen-free learning path that repeatedly revisits the same AI ideas at rising complexity, using classification as the bridge between mathematics and AI. The path moves through an ice-breaking questionnaire, a machine-learning exercise with labeled images, a fictional monster-family classification task, mushroom and fish classification rendered as Euler-Venn diagrams, tables, and decision trees, and a floor-sized decision tree that pupils physically walk. The central working objects are classification rules, feature description, and data representation; these carry the argument that AI's data-handling core is teachable to children without computers.

What would settle it

A comparison class of similar fifth graders who take the same seven-exercise post-test without receiving any of the four modules would settle the causal claim; if they score as well as the taught class, the reported improvements are not caused by the curriculum.

Watch

Extended reading notes

Core claim

The central claim is that a deliberately structured sequence of unplugged activities, anchored in the mathematics of classification and data handling, produces measurable gains in foundational AI literacy in fifth graders. The authors report that on the post-test most students could identify AI errors, use AI terminology, classify objects by features, and support their reasoning with justifications; the share of valid justifications on the Animal Footprint task rose from 9.5% on the initial questionnaire to 56.52% after instruction. They also report high enjoyment and perceived ease, with students especially responsive to activities that linked AI to real-world reasoning. In the authors' framing, the approach is a counterweight to tool-based AI education, which they argue risks teaching procedural familiarity without conceptual understanding.

Load-bearing premise

The load-bearing assumption is that post-test scores reflect learning caused by the curriculum, even though there was no control group and only one of the seven exercises had a pre-test baseline.

Editorial extensions

If this is right

  • Schools with limited technology can still deliver foundational AI concepts, since the whole path runs without computers.
  • Mathematics teachers can reinforce classification, set reasoning, and frequency interpretation while teaching AI literacy.
  • The reported increase in valid justifications on the Animal Footprint task, from 9.5% to 56.52%, indicates that a short intervention can move children from guessing to evidence-based answers.
  • High enjoyment ratings suggest that linking AI to real-world reasoning can keep primary students engaged, a factor the paper links to a noted lack of stimulating AI activities.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the reported gains are real, the same four-module structure could be adapted to other grade levels by changing the mathematical anchor—sorting for younger students, probability and accuracy for middle school—while preserving the unplugged activities.
  • Because only one exercise has a pre-test and there is no control group, the paper itself does not establish how much of the measured improvement is caused by the curriculum; a full pre-test/post-test design with a comparison class would separate learning from maturation and novelty.
  • The emphasis on multiple representations suggests a focused experiment: replacing the Venn diagrams, tables, and decision trees with verbal explanations alone should weaken classification reasoning if semiotic representation is the active ingredient.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes a structured, unplugged learning path for foundational AI literacy in fifth-grade students, integrating AI concepts with mathematical topics such as classification, set theory, and data representation. The curriculum is organized into four modules covering AI fundamentals, classification principles, classification representations, and final assessment. The empirical study involved 31 students across two classes, with 23 completing a post-test and satisfaction questionnaire. The paper reports descriptive results suggesting improvements in terminology, logical reasoning, and evaluative skills, and high student engagement. It also provides a public repository of materials and draws on established resources such as Bebras tasks and the AI4K12 framework. The central claim is that the proposed approach effectively improves foundational AI literacy and engagement, but the evidence for this claim is limited by the absence of a control group and, for most exercises, the absence of a pre-test baseline.

Significance. If fully substantiated, the paper would provide a valuable, replicable curriculum design for integrating AI literacy into primary mathematics instruction, an area that is still underdeveloped. Strengths include the explicit design grounded in constructivist and spiral-learning frameworks, the use of external validated tasks for some assessment items, and the open availability of all instructional materials in a repository. The study also addresses an important gap by focusing on conceptual understanding rather than tool usage. However, the current evidence does not support causal claims of improvement: the only direct pre-post comparison is unpaired and lacks a control condition, and the remaining exercises measure absolute post-test performance only. The paper is therefore best situated as a feasibility or design study; with appropriate reframing and stronger analysis, its contribution to the K-12 AI education literature would be meaningful.

major comments (3)
  1. [3.2 / 3.1] The only direct pre-post evidence is the Animal Footprint exercise (Section 3.2), where the paper reports 69.57% correct answers and 56.52% valid justifications versus 9.5% in the initial questionnaire. This comparison is not valid as presented because Section 3.1 states that 31 students started but only 23 completed the post-test; no paired analysis or matching of individual IDs is reported, the denominator for the 9.5% figure is not stated, and there is no control condition. The claim of 'improved significantly' therefore rests on an unsupported comparison and should be either paired with individual-level data or removed.
  2. [2.4 / 3.2] For the remaining six post-test exercises (AI Scenarios, AI Terms, Frequencies, Beaver Structure, Eulero-Venn, and Beaver Head), there is no pre-test baseline, so absolute performance levels cannot support the abstract's causal claim of improvements in terminology, logical reasoning, and evaluative skills. Section 5 acknowledges the small one-school sample and short-term focus but does not address this attribution gap. The conclusions should be recast as descriptive outcomes or the study should be explicitly labeled a feasibility study.
  3. [3.2] The word 'significantly' is used without any inferential statistical test, effect size, or confidence interval. With a final sample of 23 students, no p-values, paired tests, or effect sizes are reported anywhere in Section 3. The term 'significantly' should be replaced by a descriptive statement, or appropriate paired statistics should be provided to support the comparison in the Animal Footprint exercise.
minor comments (5)
  1. [3.1 / Figure 4] Section 3.1 says 23 students completed both the satisfaction questionnaire and the post-test, but Figure 4 percentages (e.g., 54.17%, 29.17%, 83.33%) appear to be based on 24 respondents; please reconcile the denominators.
  2. [2.2–2.4] The spelling 'Eulero-Venn' is used inconsistently with 'Euler-Venn'; please standardize the terminology throughout.
  3. [1, 2.1, 2.3, 5] The manuscript contains several typos and awkward phrases, including 'in anonymized country' (Footnote 3), 'We retain fundamental that students experience' (Section 2.1), 'Students have formulate classification rules' (Section 2.3), 'interdisciplinariety' (Section 5), and 'in is context' (Section 5). These should be corrected.
  4. [2.4 / 3.2] The post-test scoring protocol mentions that exercises 4 and 6 assess mathematical skills, but the manuscript does not state how scores in Figures 2 and 3 were standardized across exercises; please provide the scoring rubrics or point scales, or point to their location in the repository.
  5. [2.4] The paper describes the AI Scenarios and AI Terms exercises as testing understanding of AI limitations and terminology, but no pre-test or baseline is available for these constructs; the descriptive percentages should be interpreted accordingly in the discussion.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the learning-gain claim is an empirical evaluation with external tasks, not a derivation that reduces to its own inputs.

full rationale

The paper's central claim—that the unplugged curriculum improves foundational AI literacy—is supported by an empirical post-test and satisfaction survey; it is not derived from the intervention by definition. The only pre/post instrument (Animal Footprint, Sections 2.4 and 3.2) is the same task administered before and after, which is a standard measurement design rather than a circular reduction; the outcome could plausibly have improved, stayed flat, or declined. No parameter is fitted and then relabeled as a prediction; no uniqueness theorem from the authors is invoked to force the approach; and the curriculum uses external frameworks and tasks (Bebras [2], Kangourou, AI4K12 Five Big Ideas [1,34]) rather than citing the present authors' prior work as load-bearing evidence. Section 5 frankly acknowledges the small one-school sample and short-term focus. The weaknesses noted by a skeptical reader—no control group, no paired analysis, possible rater overlap—are internal-validity concerns about causal attribution, not cases where a claimed result is equivalent to its input by construction. Under the stated criteria, this warrants score 0.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The central claims rest on pedagogical assumptions and the internal validity of the evaluation; no free parameters or invented entities are introduced.

assumptions (3)
  • domain assumption Constructivism and constructionism are the correct pedagogical foundation
    The learning path is explicitly grounded in these theories (Section 2, citing [26] and [25]); no comparison to alternative pedagogical approaches is provided.
  • domain assumption Post-test scores and the survey measure the effect of the course rather than prior ability, maturation, or the novelty of a university professor
    Most exercises have no pre-test and there is no control group, so this assumption is load-bearing for the improvement claims (Section 3.1, Section 3.2).
  • domain assumption The three expert raters provide unbiased, consistent scoring
    One rater is the professor who designed and delivered the course; inter-rater reliability is not reported, yet scores are used for all outcome claims (Section 3.1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Structured Unplugged Approach for Foundational AI Literacy in Primary Education." pith.science (2026). https://pith.science/paper/2SU3YYPH

@misc{pith2026250521398,
  author       = {Pith},
  title        = {Pith review of: A Structured Unplugged Approach for Foundational AI Literacy in Primary Education},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2SU3YYPH}},
  note         = {Machine review of arXiv:2505.21398}
}
read the original abstract

Younger generations are growing up in a world increasingly shaped by intelligent technologies, making early AI literacy crucial for developing the skills to critically understand and navigate them. However, education in this field often emphasizes tool-based learning, prioritizing usage over understanding the underlying concepts. This lack of knowledge leaves non-experts, especially children, prone to misconceptions, unrealistic expectations, and difficulties in recognizing biases and stereotypes. In this paper, we propose a structured and replicable teaching approach that fosters foundational AI literacy in primary students, by building upon core mathematical elements closely connected to and of interest in primary curricula, to strengthen conceptualization, data representation, classification reasoning, and evaluation of AI. To assess the effectiveness of our approach, we conducted an empirical study with thirty-one fifth-grade students across two classes, evaluating their progress through a post-test and a satisfaction survey. Our results indicate improvements in terminology understanding and usage, features description, logical reasoning, and evaluative skills, with students showing a deeper comprehension of decision-making processes and their limitations. Moreover, the approach proved engaging, with students particularly enjoying activities that linked AI concepts to real-world reasoning. Materials: https://github.com/tail-unica/ai-literacy-primary-ed.

Figures

Figures reproduced from arXiv: 2505.21398 by the authors.

Figure 1
Figure 1. Collaborative classification task [Module 3]. [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. [RQ1] Performance distribution on AI-related post-test exercises. (a) Ex 4: Frequencies (b) Ex 6: Eulero-Venn [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. [RQ2] Performance distribution on math-related post-test exercises [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: [RQ3] Student answers about perceptions of engagement [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

41 extracted references · 39 canonical work pages

  1. [1]

    Foundational AI Literacy in Primary Education 13

    https://ai4k12.org/ (Last access 19/02/2025). Foundational AI Literacy in Primary Education 13

    work page 2025

  2. [2]

    Bebras. 2020. bebras.org. (Last access 19/02/2025)

    work page 2020

  3. [3]

    Tim Bell, Jason Alexander, Isaac Freeman, and Mick Grimley. 2009. Computer Science Unplugged: school students doing real computing without computers. New Zealand Journal of Applied Computing and Information Technology 13, 1 (2009), 20–29

    work page 2009

  4. [4]

    Bruner, J. S. (1960). The Process of Education. Harvard University Press

    work page 1960

  5. [5]

    Matteo Baldoni, Cristina Baroglio, Monica Bucciarelli, Sara Capecchi, Elena Gan- dolfi, Cristina Gena, Francesco Ianì, Elisa Marengo, Roberto Micalizio, Amon Rapp, Ivan Nabil Ras, Does Any AI-Based Activity Contribute to Develop AI Conception? A Case Study with Italian Fifth and Sixth Grade Classes, TheThirty-Eighth AAAI Conference on Artificial Intellige...

    work page

  6. [6]

    Matteo Baldoni, Cristina Baroglio, Monica Bucciarelli, Sara Capecchi, Elena Gan- dolfi, Francesco Ianì, Elisa Marengo, Roberto Micalizio, Thinking Strategies Train- ing to Support the Development of Machine Learning Understanding, A study tar- geting fifth-grade children, ICIEI 2024, April 12–14, 2024, Verbania, Italy

    work page 2024

  7. [7]

    Chevallard, La Transposition didactique: Du savoir savant au savoir enseigné, Grenoble, La Pensée sauvage, 1991 (1re éd

    Y. Chevallard, La Transposition didactique: Du savoir savant au savoir enseigné, Grenoble, La Pensée sauvage, 1991 (1re éd. 1985), 126 p. (ISBN 9782859190781)

    work page 1991

  8. [8]

    AI for Oceans

    code.org–AI and Machine Learning. AI for Oceans. 2023. https://studio.code.org/s/oceans/lessons/1/levels/6?lang=en-US (Last access 19/02/2025)

    work page 2023

Show all 41 references
  1. [9]

    CS Unplugged. [n.d.]. Principles. https://csunplugged.org/en/principles/

  2. [10]

    Dewey, J. (1938). Experience and Education. Macmillan

    1938

  3. [11]

    Digital Education Action Plan

    European Commission 2021-2027. Digital Education Action Plan. https://education.ec.europa.eu/focus-topics/ digital-education/action-plan (Last access 19/02/2025)

    2021

  4. [12]

    Stephen Frezza, Mats Daniels, Arnold Pears, Åsa Cajander, Viggo Kann, Aman- preet Kapoor, Roger McDermott, Anne-Kathrin Peters, Mihaela Sabin, and Charles Wallace. 2018. Modelling competencies for computing education beyond 2020: a research based approach to defining competenc...

    2018

  5. [13]

    Teachable Machine

    Google-Teachable machines. Teachable Machine. https://teachablemachine.withgoogle.com. (Last access 19/02/2025)

    2025

  6. [14]

    Pacheco, Matheus F

    Christiane Gresse von Wangenheim, Jean CR Hauck, Fernando S. Pacheco, Matheus F. Bertonceli Bueno. 2021. Visual tools for teaching machine learning in K- 12: A ten-year systematic mapping. Education and Information Technologies, 26(5), pp.5733-5778

    2021

  7. [15]

    Andreas Grillenberger Ralf Romeike, About Classes and Trees: Introducing Sec- ondary School Students to Aspects of Data Mining, November 2019 Lecture Notes in Computer Science In book: Informatics in Schools. New Ideas in School Informatics, 12th International Conference on In...

    2019

  8. [16]

    Informatics for All. 2023. Mission. Informatics for All. https://www.informaticsforall.org/members/ (Last access 19/02/2025)

    2023

  9. [17]

    Thinking Technology, Toward a Constructivist Design Model

    Jonassen D.H., 1994. Thinking Technology, Toward a Constructivist Design Model. Educational Technology, Vol. 34 N. 4, Pp.34-37

    1994

  10. [18]

    M., Jonassen T

    Duffy, Jonassen 1992 Duffy T. M., Jonassen T. M, Constructivism and the Tech- nology of Instruction, A Conversation, Erlbaum, Hillsdale, N.J, 1992 14 M.C. Carrisi et al

    1992

  11. [19]

    Kim, S.; Jang, Y.; Kim, W.; Choi, S.; Jung, H.; Kim, S.; and Kim, H. 2021. Why and What to Teach: AI Curriculum for Elementary School. In AAAI, 15569–15576. AAAI Press

    2021

  12. [20]

    Annabel Lindner, Stefan Seegerer, Ralf Romeike. 2019. Unplugged Activities in the Context of AI. In Informatics in Schools. New Ideas in School Informatics: 12th International Conference on Informatics in Schools: Situation, Evolution, and Per- spectives, ISSEP 2019. Proceedin...

    2019

  13. [21]

    Gonzales, Fred G

    Ruizhe Ma, Ismaila Temitayo Sanusi, Vaishali Mahipal, Joseph E. Gonzales, Fred G. Martin. 2023. Developing machine learning algorithm literacy with novel plugged and unplugged approaches. In Proc. of the 54th ACM Technical Symposium on Computer Science Education V. 1 (pp. 298-304)

    2023

  14. [22]

    MIM. 2022. Piano Nazionale Scuola Digitale (Ministero dell’Istruzione e del Mer- ito). https://www.miur.gov.it/web/ guest/scuola-digitale (Last access 19/02/2025)

    2022

  15. [23]

    Machine Learning for Kids

    ML for Kids. Machine Learning for Kids. 2023. ML for Kids, https://machinelearningforkids.co.uk/. (Last access 19/02/2025)

    2023

  16. [24]

    Papert, S.; and Solomon, C. 1971. Twenty things to do with a computer. Twenty Things to Do with a Computer, 248

    1971

  17. [25]

    Constructionism

    Harel, I., Papert, S., 1991. Constructionism. Ablex Publishing. ISBN 978- 0893917869

    1991

  18. [26]

    Piaget, J., Inhelder, B. 1969. The Psychology of the Child. New York: Basic Books

    1969

  19. [27]

    Radford, L. (2006). The semiotic turn in mathematics education: A new theory of mathematical thinking, learning, and teaching. In Semiotics in Mathematics Edu- cation (pp. 1-23)

    2006

  20. [28]

    Sabuncuoglu, A. 2020. Designing One Year Curriculum to Teach Artificial Intelli- gence for Middle School. In Proceedings of the 2020 ACM Conference on Innovation and Technology in Computer Science Education, ITiCSE ’20, 96102. Association for Computing Machinery

    2020

  21. [29]

    Sanusi, Solomon S

    Ismaila T. Sanusi, Solomon S. Oyelere, Henriikka Vartiainen, Jarkko Suhonen, Markku Tukiainen. 2023. A systematic review of teaching and learning ma- chine learning in K-12 education. Education and Information Technologies, 28(5), pp.5967-5997

    2023

  22. [30]

    Gilad Shamir, Ilya Levin, Teaching machine learning in elementary school, Inter- national Journal of Child-Computer Interaction Volume 31, March 2022, 100415,

    2022

  23. [31]

    Tanmay Sinha, Manu Kapur, Robert West, Michele Catasta, Matthias Hauswirth, and Dragan Trninic. 2020. Differential benefits of explicit failure-driven and success- driven scaffolding in problem-solving prior to instruction. Journal of Educational Psychology (2020). https://doi...

    2020 doi

  24. [32]

    Skemp, R. R. (1976). Relational understanding and instrumental understanding. Mathematics Teaching, 77, 20-26

    1976

  25. [33]

    Pages 948 - 954 https://doi.org/10.1145/3287324.3287392

    Sulmont E., Paritsas E., Cooperstock J.R., Can You Teach Me To Machine Learn?, SIGCSE ’19: Proceedings of the 50th ACM Technical Symposium on Computer Science Education. Pages 948 - 954 https://doi.org/10.1145/3287324.3287392

  26. [34]

    Touretzky, D.; Gardner-McCune, C.; Martin, F.; and Seehorn, D. 2019. Envision- ing AI for K-12: What should every child know about AI? 33rd AAAI Conference on Artificial Intelligence, AAAI 2019, 31st Innovative Applications of Artificial In- telligence Conference, IAAI 2019 an...

    2019

  27. [35]

    David Touretzky, Christina Gardner-McCune, Deborah Seehorn. 2023. Machine learning and the five big ideas in AI. International Journal of Artificial Intelligence in Education, 33(2), pp.233-266 Foundational AI Literacy in Primary Education 15

    2023

  28. [36]

    UNICEF. 2019. Workshop Report: AI and Child Rights Policy

    2019

  29. [37]

    United States Government. 2023. National Artificial Intelligence Initiative: Over- seeing and Implementing the United States National AI Strategy. AI GOV. https://www.ai.gov/ (Last access 19/02/2025)

    2023

  30. [38]

    C.; Schaper, M.-M.; Tamashiro, M.; Bilstrup, K.-E.; Lunding, M.; Graves Pe tersen, M.; and Sejer Iversen, O

    Van Mechelen, M.; Smith, R. C.; Schaper, M.-M.; Tamashiro, M.; Bilstrup, K.-E.; Lunding, M.; Graves Pe tersen, M.; and Sejer Iversen, O. 2023. Emerging Technolo- gies in K12 Education: A Future HCI Research Agenda. ACMTrans. Comput.-Hum. Interact., 30(3)

    2023

  31. [39]

    L. S. Vygotskji, Pensiero e linguaggio, Giunti, Firenze, 1966

    1966

  32. [40]

    My doll says it’s ok

    Randi Williams, Christian V. Machado, Stefania Druga, Cynthia Breazeal, Pattie Maes, "My doll says it’s ok": a study of children’s conformity to a talking doll, in Proc. of the 17th ACM Conference on Interaction Design and Children, 2018, p.625-631

    2018

  33. [41]

    Yang, W. 2022. Artificial Intelligence education for young children: Why, what, and how in curriculum design and implementation. Computers and Education: Artificial Intelligence, 3

    2022

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.