Pith. sign in

REVIEW 4 major objections 5 minor 71 references

Can the current trends of AI handle a full course of mathematics?

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper claims that ChatGPT-4o can organize an entire college math course but cannot replace the human instructor, because blind expert raters preferred human answers on 12 of 16 student questions.

desk verdict A useful but overreaching pilot: the data show GPT-4o under normal prompting lags human instructors in explanation and empathy, not that AI inherently cannot provide those things. read the letter →

arxiv 2507.21664 v1 pith:4WUVQOMZ submitted 2025-07-29 cs.AI cs.HCmath.HO

Mariam Alsayyad , Fayadh Kadhem This is my paper · ORCID
classification cs.AIcs.HCmath.HO
keywords mathematicseducationChatGPT-4ocoursesyllabuspresentationstudentquestionsassessmentdesignhuman–AIcomparisonempathyinteaching
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

This paper asks whether ChatGPT-4o, taken as a representative of current trends, can take on an entire first-year college mathematics course: writing the syllabus, presenting a topic, answering sixteen real student questions, and building a midterm exam. The authors ran a human-instructor version of the same four tasks in parallel, showed both versions to roughly thirty mathematics educators, researchers, and related specialists without identifying which was which, and compared ratings and comments. The result is a split verdict: AI is judged equal or better on organization, course learning outcomes, assessment structure, and clarity, while humans win clearly on presentations and on twelve of the sixteen student questions. The paper concludes that AI can generate accurate and well-organized course components, but it cannot supply the empathy, level-appropriate detail, and hidden emotional connection that teaching mathematics still requires, so the final course must be produced or revised by a human.

What carries the argument

The load-bearing mechanism is a paired work-sample experiment with blind evaluation. For each of four course components (syllabus, presentation, question answering, assessment), the authors created a human version and a ChatGPT-4o version of the same task, then distributed three questionnaires to about thirty expert respondents without revealing which version was human and which was AI. Quantitative ratings on five-point scales and percentage preferences, combined with qualitative comments, are used to attribute differences in quality to the human or the AI. This design is what allows the paper to separate what AI can do well (organize, integrate technology, match learning outcomes) from what it cannot (empathy, level-appropriate explanation, step-by-step intuition).

What would settle it

A preregistered, blinded replication with a larger and more diverse panel of math instructors and students, applying a statistical test to the same 16-question comparison, would falsify the central claim if it found no significant preference for the human answers or a preference for the AI answers.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central finding is that the current generation of AI, specifically ChatGPT-4o, can handle the organizational and accurate-computation components of a full college mathematics course but cannot handle the interpersonal component. In blind expert evaluation, the AI syllabus scored higher on learning outcomes, assessment design, and organization, and the AI exam was rated clearer and better aligned with course outcomes; the human presentation outscored the AI's on virtually every criterion, and human answers to real student questions were preferred on 12 of 16 questions, with the highest human preference at 72.4% versus the AI's highest at 50%. Respondents described the human answers as empathetic, detailed, intuitive, and matched to the students' level, whereas AI answers were described as abstract, short, and pitched at specialists. The authors therefore claim that there is 'a hidden emotional part, even in science, that cannot be fulfilled by the AI in its current state', and recommend a division of labor: humans provide the creative ideas, weekly structure, level calibration, empathy, and final revision, while AI contributes speed, organization, technology integration, and clarity.

Load-bearing premise

The study depends on the assumption that the preference judgments of about thirty respondents, many of them Bahraini colleagues and friends of the authors who filled the questionnaire, are a valid and unbiased measure of what makes a full mathematics course work, and that those respondents genuinely stayed blind to which materials were AI-generated.

Editorial extensions

If this is right

  • AI-generated syllabi and assessments can serve as strong first drafts, with humans supplying the weekly breakdown, workload balance, and logical sequencing that respondents preferred in the human version.
  • The 16-question results imply that for routine student questions in early calculus, AI answers are not yet adequate at the students' actual level; they need human rewriting or detailed prompting that specifies the audience.
  • The near-even split on assessments suggests that AI can already generate a usable midterm exam, but the grading distribution needs an instructor's check because 11.1% of respondents judged the AI's grade distribution unfair.
  • The paper's recommended workflow—human ideas and structures, AI drafting, human revision—implies that institutions adopting AI in math teaching should keep teaching staff in the loop rather than replacing them.

Reading between the lines

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

  • The same blind-comparison method could be extended to students as participants; the authors list this as future work, and student preferences might differ from expert preferences, especially on whether shorter AI explanations aid or hinder learning.
  • If the finding generalizes, the durable question is whether the emotional and level-appropriate component of teaching can eventually be captured by richer prompting, which would move the boundary between human and AI in math teaching over time.
  • The sample is small, local, and partly composed of the authors' colleagues, so the result is best read as a case study of current GPT-4o behavior in one institutional context rather than a universal law.
  • A direct test of the mechanism would be to feed the human answers as few-shot examples into GPT-4o and check whether the 12-of-16 human preference flips; if it does, the deficit is a prompt/context problem rather than a fundamental emotional blind spot.
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

4 major / 5 minor

Summary. The paper reports a comparative study of human-prepared course materials (syllabus, lecture slides, answers to 16 student questions, and a midterm exam) against materials generated by ChatGPT-4o under a 'normal use' prompting protocol. Around 30 mathematics instructors and related professionals completed questionnaires comparing the anonymous materials, and the authors analyzed both quantitative ratings and open-ended comments. The paper finds that the AI was competitive or superior on organizational aspects (syllabus structure, assessment clarity) but that respondents preferred the human materials on presentation flow, detail, and responsiveness to students' level and emotions. It concludes that AI cannot, in its current state, supply the "hidden emotional part" of mathematics teaching, and it recommends a hybrid workflow in which humans supply creative and emotional direction and AI assists with drafting and organization.

Significance. If taken as an exploratory case study, the paper offers a useful, transparent artifact: the full set of 16 question/answer pairs is reproduced in an appendix, the prompting choice is explicitly justified as representing normal instructor use, and the authors are candid about time savings and about the need for human revision. This makes the work a potential teaching resource and a modest empirical data point for AI-in-education discussions. However, the significance of the central claim is limited by the small convenience sample, the single model and single prompt condition, the absence of inferential statistics, and the authors' dual role as creators of the human materials and analysts of the responses. The conclusion as stated overreaches the evidence, though the reported preference patterns are plausible and worth reporting if properly qualified.

major comments (4)
  1. [Abstract; §3 (AI system choice); §6.3] The central conclusion that there is an emotional part that "cannot be fulfilled by the AI in its current state" is a capability claim, but the experiment is a single-condition performance observation. The paper deliberately used "normal use" prompting without empathy-focused instructions, and then in §6.3 it states that AI answers "could be improved by requesting that students' emotions, age, and previous knowledge be considered, and a caring and compassionate communication tone should be used." If explicit prompting can add these qualities, the observed deficit is a property of the chosen prompt, not an inherent limitation of "the AI in its current state." To support the "cannot" wording, the study would need at least one additional condition with an empathy- and audience-aware prompt, or the conclusion must be restricted to the tested prompting protocol. This is not a minor wording issue; it affects the paper's core message.
  2. [§3 (Respondents); §4; §5.3] The quantitative support is thin for the strength of the claims. The study relies on approximately 30 respondents, many of them Bahraini colleagues and friends of the authors, with no reported response rate, no item-level Ns, no confidence intervals, and no inferential statistics. For example, §4.3 reports that 54.55% of respondents preferred the human-generated assessment and 45.45% the AI one, which §5.3 correctly notes "differs by 2 respondents only." Similar differences in the syllabus and presentation ratings (e.g., 4.07 vs. 3.63 out of 5 in §4.1) are reported without error bars or significance tests. The paper should either report exact denominators and appropriate statistical tests, or explicitly frame all comparative statements as descriptive tendencies from a small convenience sample rather than as findings that support the general "cannot" claim.
  3. [§3 (Methodology); §5 (Qualitative Results)] The authors created the human materials, designed the questionnaires, and also analyzed the open-ended comments. No coding scheme, second coder, inter-rater reliability metric, or audit trail is reported for the qualitative analysis, and the authors were not blind to which materials were human-generated when interpreting the comments. This double role creates a real risk that the qualitative narrative overweights supportive comments and interprets ambiguous remarks in favor of the human-authored materials. The manuscript should acknowledge this as a limitation and should provide at least a summary of the coding categories and the raw comment excerpts, or use an independent analyst for the qualitative coding.
  4. [§4.2; Appendix] The presentation of the question-level results is incomplete. The text says that out of 16 questions, 12 favored the human answers, one favored AI, two had equal analysis, and one had an equal split, but the appendix percentages (e.g., Q16: human 42.9%, AI 50%) show that the complement is split between "equal analysis" and "neither" without giving the actual counts or the total number of respondents per question. Without per-item denominators, the reader cannot assess whether a 72.4% vs. 17.9% gap is based on 29 responses or on 14 responses. Please provide the full response counts for every option in the appendix table, or a supplementary data file.
minor comments (5)
  1. [§5.2; §6.4; §7; Appendix] There are several typographical errors that should be corrected: "worthening" (§5.2), "reveled" (§6.4), "questionnaireing" (§7), "grater" (appendix Q16), and inconsistent spacing in expressions like "∞ − ∞and ∞ ∞" in the abstract and appendix.
  2. [§6.3] The sentence "Therefore, by providing comprehensive responses and expanding them with examples that are suitable for this age group of students, AI's answers could be used in the classroom" appears twice in the same subsection; one occurrence should be deleted.
  3. [§3 (Methodology); §4.3] The number of respondents is reported inconsistently: the text says "approximately 30 respondents" but the assessment results use percentages that imply totals around 27 (for the clarity ratings) and around 22 (for the final preference). Please report the exact number of respondents for each questionnaire and for each item, since the current presentation makes it difficult to assess the precision of the percentages.
  4. [References] The reference list uses inconsistent formatting (e.g., missing volume/page ranges for [1], [7], [16], and [47]) and some entries lack DOIs. A consistent citation style would improve readability.
  5. [§6.2] The time comparison "the presentation created by the authors took an average of 3 hours per day, and it was done in 4 days, while the change requests for the AI were done in counted seconds" is an interesting practical observation, but it would be clearer to state the total human hours (about 12) rather than an average per day, since the latter depends on an arbitrary day length.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper is an empirical comparison with an inductive conclusion, not a derivation whose output is preloaded into its inputs.

full rationale

This manuscript reports a questionnaire-based comparison between human-prepared course materials and GPT-4o-generated materials; it contains no formal derivation chain, no fitted parameters, and no prediction that is computed from the data it is meant to explain. The central claim that 'there is still a hidden emotional part... that cannot be fulfilled by the AI in its current state' is an inductive generalization from respondent preferences and qualitative comments, not a quantity defined in terms of those same preferences. The reference list contains no prior work by Alsayyad or Kadhem, so no load-bearing self-citation is present, and no uniqueness theorem or ansatz is imported from the authors' own earlier papers. The nearest concerns are methodological rather than circular: the authors created the 'human' materials they then evaluated, the respondent sample is small and largely local, and the 'cannot' conclusion is stronger than a single-prompt, single-model comparison supports. These are validity and overreach risks, not cases where an equation or result reduces by construction to its own input. Therefore no specific circular step can be exhibited under the required standard.

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

The paper introduces no free parameters and no invented entities. Its load-bearing assumptions are the representativeness of GPT-4o, the validity of expert preference judgments, and the fairness of the authors' self-generated human baseline.

assumptions (3)
  • domain assumption GPT-4o represents the 'current trends of AI' for teaching math
    Section 3 'AI system choice' states that ChatGPT is the most trending tool and is used as a stand-in for AI generally; this assumption limits all conclusions to one commercial chatbot.
  • domain assumption Expert questionnaire judgments are a valid measure of course quality
    The paper treats the preferences of roughly 30 respondents as evidence about learning outcomes without validating the questionnaire or testing inter-rater reliability.
  • domain assumption The authors' own course materials fairly represent 'human' work
    The human side of the comparison was produced by the authors, as stated in Section 3, so the result reflects two specific instructors, not a general human baseline.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Can the current trends of AI handle a full course of mathematics?." pith.science (2026). https://pith.science/paper/4WUVQOMZ

@misc{pith2026250721664,
  author       = {Pith},
  title        = {Pith review of: Can the current trends of AI handle a full course of mathematics?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4WUVQOMZ}},
  note         = {Machine review of arXiv:2507.21664}
}
read the original abstract

This paper addresses the question of how able the current trends of Artificial Intelligence (AI) are in managing to take the responsibility of a full course of mathematics at a college level. The study evaluates this ability in four significant aspects, namely, creating a course syllabus, presenting selected material, answering student questions, and creating an assessment. It shows that even though the AI is strong in some important parts like organization and accuracy, there are still some human aspects that are far away from the current abilities of AI. There is still a hidden emotional part, even in science, that cannot be fulfilled by the AI in its current state. This paper suggests some recommendations to integrate the human and AI potentials to create better outcomes in terms of reaching the target of creating a full course of mathematics, at a university level, as best as possible.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

71 extracted references · 70 canonical work pages

  1. [1]

    Acceptance of artificial intelligence in teaching science: Science teachers’ perspective

    Al Darayseh, A., 2023. Acceptance of artificial intelligence in teaching science: Science teachers’ perspective. Computers and Education: Artificial Intelligence, 4, p.100132

    work page 2023

  2. [2]

    and Ajjawi, R., 2023

    Bearman, M., Nieminen, J.H. and Ajjawi, R., 2023. Designing assessment in a digital world: An organising framework. Assessment & Evaluation in Higher Education, 48(3), pp.291–304

    work page 2023

  3. [3]

    and Ruiz-Herrera, S., 2020

    Castrill´ on, O.D., Sarache, W. and Ruiz-Herrera, S., 2020. Predicci´ on del rendimiento acad´ emico por medio de t´ ecnicas de inteligencia artificial.Formaci´ on Universitaria,13, pp.93–102

    work page 2020

  4. [4]

    and Bhattacharjee, K.K., 2020

    Chatterjee, S. and Bhattacharjee, K.K., 2020. Adoption of artificial intelligence in higher education: A quantitative analysis using structural equation modelling. Education and Information Technologies, 25, pp.3443–3463

    work page 2020

  5. [5]

    and Chua, X.N., 2020

    Chew, E. and Chua, X.N., 2020. Robotic Chinese language tutor: Personalising progress assessment and feedback or taking over your job? Horizont, 28, pp.113–124

    work page 2020

  6. [6]

    and Alper, A., 2024

    Coskun, T.K. and Alper, A., 2024. Evaluating the evaluators: A comparative study of AI and teacher assessments in Higher Education. Digital Education Review, (45), pp.124- 140. 21

    work page 2024

  7. [7]

    Effective learning strategies and Artificial Intelligence (AI) support for accelerated math acquisition

    Cunska, A., 2020. Effective learning strategies and Artificial Intelligence (AI) support for accelerated math acquisition. In: European Proceedings of International Conference on Education and Educational Psychology. European Publisher

    work page 2020

  8. [8]

    and Berner, J., 2023.Mathematical capabilities of ChatGPT.arXiv preprint arXiv:2301.13867

    Frieder, S., Pinchettil, L., Griffiths, R., Salvatori, T., Lukasiewicz, T., Petersen, P., Chevalier, A. and Berner, J., 2023.Mathematical capabilities of ChatGPT.arXiv preprint arXiv:2301.13867

    arXiv 2023

Show all 71 references
  1. [9]

    The future of learning: AI-based curriculum development

    Ejjami, R., 2024. The future of learning: AI-based curriculum development. International Journal for Multidisciplinary Research, 6(4), pp.1–31

    2024

  2. [10]

    and Tang, X., 2023

    Gamage, K.A., Dehideniya, S.C., Xu, Z. and Tang, X., 2023. ChatGPT and higher education assessments: More opportunities than concerns?. Journal of Applied Learning and Teaching, 6(2), pp.358-369

    2023

  3. [11]

    and Roig-Vila, R., 2021

    Gonz´ alez-Calatayud, V., Prendes-Espinosa, P. and Roig-Vila, R., 2021. Artificial intel- ligence for student assessment: A systematic review. Applied sciences, 11(12), p.5467

    2021

  4. [12]

    Graide, 2022. Graide. https://www.graide.co.uk/

    2022

  5. [13]

    AI and education: The importance of teacher and student relations

    Guilherme, A., 2019. AI and education: The importance of teacher and student relations. AI & SOCIETY, 34(1), pp.47–54

    2019

  6. [14]

    and Liu, S., 2023

    Hannan, E. and Liu, S., 2023. AI: new source of competitiveness in higher education. Competitiveness Review: An International Business Journal, 33(2), pp.265–279

    2023

  7. [15]

    and Lee, A., 2024, July

    Henkel, O., Horne-Robinson, H., Kozhakhmetova, N. and Lee, A., 2024, July. Effective and scalable math support: Experimental evidence on the impact of an AI-math tutor in Ghana. In International Conference on Artificial Intelligence in Education (pp. 373-381). Cham: Springer N...

    2024

  8. [16]

    and Tuomi, I., 2022

    Holmes, W. and Tuomi, I., 2022. State of the art and practice in AI in education. European Journal of Education, 57(4), pp.542–570

    2022

  9. [17]

    and Hossain, M.S., 2022

    Hooda, M., Rana, C., Dahiya, O., Rizwan, A. and Hossain, M.S., 2022. Artificial in- telligence for assessment and feedback to enhance student success in higher education. Mathematical Problems in Engineering, 2022(1), p.5215722

    2022

  10. [18]

    and Nerdel, C., 2023

    Hornberger, M., Bewersdorff, A. and Nerdel, C., 2023. What do university students know about Artificial Intelligence? Development and validation of an AI literacy test. Computers and Education: Artificial Intelligence, 5, p.100165

    2023

  11. [19]

    and Liu, G., 2022

    Hou, J., Li, Z. and Liu, G., 2022. Macro education approach to improve learning interest under the background of artificial intelligence. Wireless Communications and Mobile Computing, 2022, Article 4295887. 22

    2022

  12. [20]

    and B´ elanger, J.J., 2023

    Ibrahim, H., Liu, F., Asim, R., Battu, B., Benabderrahmane, S., Alhafni, B., Adnan, W., Alhanai, T., AlShebli, B., Baghdadi, R. and B´ elanger, J.J., 2023. Perception, per- formance, and detectability of conversational artificial intelligence across 32 university courses. Scie...

    2023

  13. [21]

    Rethinking online assessment strategies: Authenticity versus AI chatbot intervention

    Ifelebuegu, A., 2023. Rethinking online assessment strategies: Authenticity versus AI chatbot intervention. Journal of Applied Learning and Teaching, 6(2), pp.1–14

    2023

  14. [22]

    Artificial intelligence in education: The urgent need to prepare teach- ers for tomorrow’s schools

    Karsenti, T., 2019. Artificial intelligence in education: The urgent need to prepare teach- ers for tomorrow’s schools. Formation et Profession [Education and Profession], 27(1), pp.112–116

    2019

  15. [23]

    and Kasneci, G., 2023

    Kasneci, E., Sessler, K., K¨ uchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh, G., G¨ unnemann, S., H¨ ullermeier, E., Krusche, S., Kutyniok, G., Michaeli, T., Nerdel, C., Pfeffer, J., Poquet, O., Sailer, M., Schmidt, A., Seidel, T. and Kasneci, G., 202...

    2023

  16. [24]

    and Kim, M.K., 2022, March

    Kim, N.J. and Kim, M.K., 2022, March. Teacher’s perceptions of using an artificial intelligence-based educational tool for scientific writing. In Frontiers in education (Vol. 7, p. 755914). Frontiers Media SA

    2022

  17. [25]

    and Kim, J.-H., 2020

    Kim, W.-H. and Kim, J.-H., 2020. Individualized AI tutor based on developmental learn- ing networks. IEEE Access, 8, pp.27927–27937

    2020

  18. [26]

    and Liu, T., 2024

    Liang, S., Zhang, W., Zhong, T. and Liu, T., 2024. Mathematics and machine creativity: A questionnaire on bridging mathematics with AI . arXiv preprint arXiv:2412.16543

    2024

  19. [27]

    and Chocksathaporn, P., 2023

    Limna, P., Kraiwanit, T., Jangjarat, K., Klayklung, P. and Chocksathaporn, P., 2023. The use of ChatGPT in the digital era: Perspectives on chatbot implementation. Journal of Applied Learning and Teaching, 6(1), pp.64–74

    2023

  20. [28]

    and Qin, J., 2021

    Lin, P.Y., Chai, C.S., Jong, M.S.Y., Dai, Y., Guo, Y. and Qin, J., 2021. Modelling the structural relationship among primary students’ motivation to learn artificial intel- ligence. Computers and Education: Artificial Intelligence, 2, p.100006

    2021

  21. [29]

    and Yang, S., 2023

    Lin, C., Huang, A. and Yang, S., 2023. Review of AI-driven conversational chatbots implementation methodologies and challenges (1999–2022). Sustainability, 15, p.4012

    2023

  22. [30]

    and Liu, L., 2017

    Liu, M., Wang, Y., Xu, W. and Liu, L., 2017. Automated scoring of Chinese engineering students’ English essays. International Journal of Distance Education Technologies, 15, pp.52–68

    2017

  23. [31]

    Artificial intelligence applications: An introduction to education development in the light of coronavirus pandemic COVID-19 challenges

    Mahmoud, A., 2020. Artificial intelligence applications: An introduction to education development in the light of coronavirus pandemic COVID-19 challenges. International Journal of Research in Educational Sciences, 3(4), pp.171–224. 23

    2020

  24. [32]

    and Tan, S., 2023

    Rudolph, J., Tan, S. and Tan, S., 2023. ChatGPT: Bullshit spewer or the end of tradi- tional assessments in higher education?. Journal of applied learning and teaching, 6(1), pp.342-363

    2023

  25. [33]

    and Garro-Aburto, L.L., 2019.In- teligencia artificial y sus implicaciones en la educaci´ on superior

    Oca˜ na-Fern´ andez, Y., Valenzuela-Fern´ andez, L.A. and Garro-Aburto, L.L., 2019.In- teligencia artificial y sus implicaciones en la educaci´ on superior. Prop´ ositos y Repre- sentaciones, 7

    2019

  26. [34]

    ChatGPT (July 2025 version) [online]

    OpenAI, 2025. ChatGPT (July 2025 version) [online]. Available at: https://chat. openai.com/ [Accessed 27 Jul. 2025]

    2025

  27. [35]

    and Lan, A., 2024

    Otero, N., Druga, S. and Lan, A., 2024. A Benchmark for Math Misconceptions: Bridging Gaps in Middle School Algebra with AI-Supported Instruction. arXiv preprint arXiv:2412.03765

    2024 arXiv

  28. [36]

    and Bassey, B.A., 2023

    Owan, V.J., Abang, K.B., Idika, D.O., Etta, E.O. and Bassey, B.A., 2023. Exploring the potential of artificial intelligence tools in educational measurement and assessment. EURASIA Journal of Mathematics, Science and Technology Education, 19(8), p.em2307

    2023

  29. [37]

    and Wylie, R., 2016

    Roll, I. and Wylie, R., 2016. Evolution and revolution in artificial intelligence in educa- tion. International journal of artificial intelligence in education, 26(2), pp.582-599

    2016

  30. [38]

    and Akbar, A., 2025

    Samee Ullah, R., Hashim, F., Bandeali, M.M. and Akbar, A., 2025. Artificial Intelligence in Curriculum Design: A Roadmap for Adaptive and Personalized Learning in Higher Education. The Critical Review of Social Sciences Studies, 3(3), pp.304–322

    2025

  31. [39]

    and Vega-Fleitas, E., 2023

    S´ anchez-Ruiz, L.M., Moll-L´ opez, S., Nu˜ nez-P´ erez, A., Mora˜ no-Fern´ andez, J.A. and Vega-Fleitas, E., 2023. ChatGPT challenges blended learning methodologies in engineer- ing education: A case study in mathematics. Applied Sciences, 13(10), p.6039

    2023

  32. [40]

    Should robots replace teachers?: AI and the future of education

    Selwyn, N., 2019. Should robots replace teachers?: AI and the future of education. John Wiley & Sons

    2019

  33. [41]

    and K¨ arkk¨ ainen, T., 2025, March.The Use of Gen- erative Artificial Intelligence for Upper Secondary Mathematics Education Through the Lens of Technology Acceptance

    Set¨ al¨ a, M., Heilala, V., Sikstr¨ om, P. and K¨ arkk¨ ainen, T., 2025, March.The Use of Gen- erative Artificial Intelligence for Upper Secondary Mathematics Education Through the Lens of Technology Acceptance. In Proceedings of the 40th ACM/SIGAPP Symposium on Applied Compu...

    2025

  34. [42]

    and Pandian, S.S., 2020

    Somasundaram, M., Latha, P. and Pandian, S.S., 2020. Curriculum design using ar- tificial intelligence (AI) back propagation method. Procedia Computer Science, 172, pp.134–138

    2020

  35. [43]

    and Thomas, A., 2023

    Southworth, J., Migliaccio, K., Glover, J., Glover, J.N., Reed, D., McCarty, C., Bren- demuhl, J. and Thomas, A., 2023. Developing a model for AI Across the curriculum: Transforming the higher education landscape via innovation in AI literacy. Computers and Education: Artifici...

    2023

  36. [44]

    and Chu, S.K.W., 2023

    Su, J., Ng, D.T.K. and Chu, S.K.W., 2023. Artificial intelligence (AI) literacy in early childhood education: The challenges and opportunities. Computers and Education: Ar- tificial Intelligence, 4, p.100124

    2023

  37. [45]

    and Neag, R., 2020

    Szabo, Z.K., K¨ ortesi, P., Guncaga, J., Szabo, D. and Neag, R., 2020. Examples of problem-solving strategies in mathematics education supporting the sustainability of 21st- century skills. Sustainability, 12, p.10113

    2020

  38. [46]

    ‘Heartware’ for the compassionate teacher: Humanizing the academy through mindsight, attentive love, and storytelling

    Tan, E., 2022. ‘Heartware’ for the compassionate teacher: Humanizing the academy through mindsight, attentive love, and storytelling. Journal of Applied Learning & Teach- ing, 5(2), pp.152–159

    2022

  39. [47]

    and Berman, A., 2023

    Tenakwah, E.S., Boadu, G., Tenakwah, E.J., Parzakonis, M., Brady, M., Kansiime, P., Said, S., Ayilu, R., Radavoi, C. and Berman, A., 2023. Generative AI and higher education assessments: A competency-based analysis . Research Square. Available at: https://doi.org/10.21203/rs.3...

    2023 doi

  40. [48]

    and P´ erez-Suay, A., 2022.A classification of artificial intelligence systems for mathematics education

    Van Vaerenbergh, S. and P´ erez-Suay, A., 2022.A classification of artificial intelligence systems for mathematics education. In Mathematics education in the age of artificial intelligence: How artificial intelligence can serve mathematical human learning (pp. 89- 106). Cham: ...

    2022

  41. [49]

    and Palacios-Pacheco, X., 2020

    Villegas-Ch, W., Arias-Navarrete, A. and Palacios-Pacheco, X., 2020. Proposal of an architecture for the integration of a chatbot with artificial intelligence in a smart campus for the improvement of learning. Sustainability, 12, p.1500

    2020

  42. [50]

    and Jarrah, A.M., 2023

    Wardat, Y., Tashtoush, M.A., AlAli, R. and Jarrah, A.M., 2023. ChatGPT: A revo- lutionary tool for teaching and learning mathematics. Eurasia Journal of Mathematics, Science and Technology Education, 19(7), p.em2286

    2023

  43. [51]

    ‘A revolution in productivity’: What ChatGPT could mean for business

    Watkins, M.D., 2022. ‘A revolution in productivity’: What ChatGPT could mean for business. [online] IMD. Available at: <https://www.imd.org/ibyimd/technology/a- revolution-in-productivity-what-chatgpt-could-mean-for-business/> [Accessed 27 Jul. 2025]

    2022

  44. [52]

    and Chai, C.S., 2022

    Xia, Q., Chiu, T.K.F., Lee, M., Temitayo, I., Dai, Y. and Chai, C.S., 2022. A self- determination theory design approach for inclusive and diverse artificial intelligence (AI) K–12 education. Computers & Education, 189, p.104582

    2022

  45. [53]

    and Yi, H., 2021

    Xiao, M. and Yi, H., 2021. Building an efficient artificial intelligence model for per- sonalized training in colleges and universities. Computer Applications in Engineering Education, 29, pp.350–358

    2021

  46. [54]

    and Gouverneur, F., 2019

    Zawacki-Richter, O., Mar ´ ın, V.I., Bond, M. and Gouverneur, F., 2019. Systematic re- view of research on artificial intelligence applications in higher education–where are the educators?. International journal of educational technology in higher education, 16(1), pp.1-27. 25

    2019

  47. [55]

    equal analysis

    Zhao, C., 2024 AI-assisted assessment in higher education: A systematic review. Journal of Educational Technology and Innovation, Vol. 6 No. 4 (2024). 26 Appendix The following table represents the questions and answers of questionnaire 2 that was men- tioned above. The comple...

    2024

  48. [56]

    We use it to denote something that is uncontrollably big, but it is usually not a number

    Why is ∞ − ∞and ∞ ∞ not equal to 0 and 1, respectively? Human’s Answer (60.7%) AI’s Answer (32.1%) We can say that ∞ is a description more than a number. We use it to denote something that is uncontrollably big, but it is usually not a number. Now, we can write ∞ twice, but ea...

  49. [57]

    Thus, when we square something that is dramatically large, we still get something very very large and unbounded

    Why ∞2 = ∞? Should not ∞2 be bigger? Human’s Answer (37%) AI’s Answer (33.3%) The symbol ∞ in this context represents a description, not a usual number. Thus, when we square something that is dramatically large, we still get something very very large and unbounded. So, we stil...

  50. [58]

    Human’s Answer (21.4%) AI’s Answer (35.7%) The symbol x3 means we have 3 copies of x

    Why x3 x = x2? I feel they are the same as we have x above and below. Human’s Answer (21.4%) AI’s Answer (35.7%) The symbol x3 means we have 3 copies of x. Thus, x3 x = x · x · x x . Now, we can cancel out one of the x’s on the top with the one on the bottom. This will give us...

  51. [59]

    Usually, if we have polynomials that are easy to factorize, we will be able to cancel some common factors out of the expression in the numerator and denominator

    When do we use L’Hˆ opital’s rule and when do we factorize? Human’s Answer (35.7%) AI’s Answer (35.7%) Firstly, we need to substitute the limit values to check if the situation is indeterminate, for example, an ∞ ∞ , or a 0 0 . Usually, if we have polynomials that are easy to ...

  52. [60]

    This for example happens if we deal with a distance problem or something similar

    In solving a quadratic equation, should not the negative solution be excluded? Human’s Answer (39.3%) AI’s Answer (28.6%) The answer is yes and no! In fact, and this is a common case, if we have a real-life problem that only allows positive answers, we will reject the negative...

  53. [61]

    Because a limit describes the behavior of a function as it approaches a value, not necessarily its value at that point

    Why is the limit still the same even if we have a hole in the graph? Human’s Answer (17.9%) AI’s Answer (32.1%) The hole affects the value of a function at a specific point, but the limit studies the behavior of the function closer to the number from the positive and negative ...

  54. [62]

    As a result, this will give a2 + 2ab + b2, which is clearly not equal the second expression a2 + b2

    Why (a + b)2 ̸= a2 + b2? What is the difference? Human’s Answer (58.6%) AI’s Answer (20.7%) First, notice that the LHS means that we multiply the expression ( a + b) by itself twice. As a result, this will give a2 + 2ab + b2, which is clearly not equal the second expression a2...

  55. [63]

    The LHS means that we have to take the sum, then take the square root, while the second means we take the sum of the values after taking the square root

    Why √ a + b ̸= √a + √ b? What is the difference? Human’s Answer (72.4%) AI’s Answer (6.9%) It is not correct to assume that it is applicable to distribute the square root over the terms and add them to each other, as it is not distributive. The LHS means that we have to take t...

  56. [64]

    We use it to denote something that is uncontrollably big, but it is usually not a number

    Can an ∞ be larger or smaller than other ∞? Basically, what is ∞? Human’s Answer (44.8%) AI’s Answer (34.5%) We can say that ∞ is a description more than a number. We use it to denote something that is uncontrollably big, but it is usually not a number. Now, we can write ∞ twi...

  57. [65]

    The equal sign (=) is usually used to denote equations that are true for some specific values

    What is the difference between the equal (=) sign and the identity ( ≡) sign? Human’s Answer (44.8%) AI’s Answer (37.9%) First, we need to understand there are two types of equations, namely a conditional equation and an identity. The equal sign (=) is usually used to denote e...

  58. [66]

    What is equating the coefficient? Is it a fixed rule? Human’s Answer (55.2%) AI’s Answer (20.7%) In general, two polynomials are equal if and only if • their constant terms are equal, • the factors in front of x in both polynomials are equal, • the factors in front of x2 in bo...

  59. [67]

    This means that they just force this equation to be true

    How is 0! equal to 1? Human’s Answer (65.9%) AI’s Answer (10.3%) First, many textbooks say that 0! = 1, by definition. This means that they just force this equation to be true. However, if we want to see the intuition behind that, we can notice that, in general, n! = (n + 1)! ...

  60. [68]

    • 1 10 = 0.1

    Why is dividing any number by ∞ equal to zero? Human’s Answer (51.7%) AI’s Answer (17.2%) First, let us do some calculations: • 1 2 = 0.5. • 1 10 = 0.1. • 1 100 = 0.01. • 1 1000 = 0.001. • 1 10000 = 0.0001. We can notice that when the number of the denominator becomes bigger, ...

  61. [69]

    Usually, it is used when other factors can be cancelled out from the expression

    What is the difference between factorizing a3 + b3 and expanding (a + b)3? Human’s Answer (62.1%) AI’s Answer (20.7%) Factorization is used to simplify the expression as a product of its factors. Usually, it is used when other factors can be cancelled out from the expression. ...

  62. [70]

    What is the imaginary number j in a complex number? Human’s Answer (57.1%) AI’s Answer (21.4%) The imaginary number i or j, depending on the book notation, is the square root of −1, that is, √−1. In fact, the square root is only defined on nonnegative numbers, but mathematicia...

  63. [71]

    What is the difference between an expected event and the probability of an event? Human’s Answer (42.9%) AI’s Answer (50%) Probability is calculating the likelihood of a certain event to happen. The probability represents either a certainty that an event will occur, with a val...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.