Pith. sign in

REVIEW 4 major objections 5 minor 86 references

This paper establishes that the fit between generative AI and K-12 classrooms varies along a measurable spectrum of 'cultural distance,' from near-seamless alignment to adaptation that fails entirely.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review

2026-08-04 17:35 UTC pith:WV5KQTLK

load-bearing objection A useful middle-range framework with a real tautology problem at its core; worth engaging, but the taxonomy needs an independent handle before I'd fully trust the levels. the 4 major comments →

arxiv 2509.10780 v1 pith:WV5KQTLK submitted 2025-09-13 cs.HC cs.AI

Bridging Cultural Distance Between Models Default and Local Classroom Demands: How Global Teachers Adopt GenAI to Support Everyday Teaching Practices

Ruiwei Xiao , Qing Xiao , Xinying Hou , Hanqi Jane Li , Phenyo Phemelo Moletsane , Hong Shen , John Stamper This is my paper
classification cs.HC cs.AI
keywords cultural distancegenerative AI in educationK-12 teachershuman-AI value alignmentcross-cultural HCIteacher laborAI adoptionlocalization
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper argues that generative AI tools carry a 'default culture'—norms, curricula, language, and communication styles drawn from their training data—that sits at a variable distance from what any given K-12 classroom actually demands. To capture that gap, the authors define 'cultural distance' and show, from 30 interviews with teachers in South Africa, Taiwan, and the US, that it falls into three levels: low, where a quick edit makes the output usable; mid, where teachers must prompt, revise, or switch tools to get acceptable results; and high, where adaptation fails no matter how hard they try. The point of the framework is to make visible the cultural labor teachers perform after adopting GenAI and to locate where responsibility for the gap should fall—on users, designers, or policymakers. If the paper is right, 'cultural distance' becomes a transferable analytic lens for analyzing AI alignment across domains, not just education.

Core claim

On its own terms, the paper establishes that the fit between chat-based GenAI and everyday teaching is not binary (biased vs. aligned) but a spectrum, and that this spectrum can be described by six recurring categories under three levels of cultural distance. At low distance, routine tasks like stakeholder communication and brainstorming activities align with the model's default strengths, needing only minor edits. At mid distance, assessment design and culturally relevant activities demand deliberate prompting, sustained revision, or the use of education-specific tools. At high distance, tasks fail entirely: local low-resource languages produce error-laden or refused responses, and policy c

What carries the argument

The central object is the concept of 'cultural distance'—the gap between GenAI's default cultural repertoire and the situated demands of teaching practice—operationalized through the amount of effort teachers must invest to make outputs usable. The framework is carried by a qualitative analysis of 30 semi-structured interviews (10 per region in South Africa, Taiwan, and the US), from which six categories emerged, two per level of effort (low, mid, high). The effort axis is the load-bearing mechanism: it converts an abstract cultural mismatch into an observable, comparable quantity across tasks and regions.

Load-bearing premise

The load-bearing premise is that the effort a teacher reports investing is a valid, comparable measure of the distance between GenAI defaults and local classroom demands; if teachers differ in prompting skill, persistence, or tolerance for imperfection, the same underlying gap could produce different effort levels, and the three-level split would be a property of the teacher rather than the gap.

What would settle it

Track objective effort (time, number of revisions, prompt iterations) across the six task categories with teachers of matched prompting proficiency; if the low-mid-high ordering fails to reproduce, or if expert prompters bridge reported high-distance tasks (e.g., Sepedi prompts), the framework's effort axis reflects teacher skill rather than cultural distance.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • If the framework holds, teachers' adaptation work—prompting, editing, reframing, supplementing with local knowledge—becomes visible as a form of cultural labor that current GenAI design does not count.
  • Designers can use the distance level as a diagnostic: mid-distance tasks point toward curriculum-aware fine-tuning and grade-level calibration; high-distance tasks point toward training data expansion and transparent limitation messaging.
  • Policymakers and institutions can see where user effort can close the gap and where only structural change (data, infrastructure, regulation) will, shifting responsibility away from individual teachers.
  • The recurrence of the same low/mid/high pattern across three very different contexts suggests the framework is generalizable beyond any single region, as the paper claims.
  • The paper's open-ended framing invites testing the taxonomy in other professions and user groups, where the same three levels may reappear.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • [Editorial inference] The paper's own limitation section notes the 30-interview sample is modest and self-reported; building on that, the effort axis could be calibrated with logs or class observations, which would test whether the low-mid-high split is a stable property of tasks or an artifact of teacher self-assessment.
  • [Editorial inference] The effort-based operationalization is vulnerable to individual differences: a teacher with strong prompting skills may experience a mid-distance task as low-distance, so the three-level split may partly describe the teacher, not the gap. This is not addressed by the paper's self-report-only data.
  • [Editorial inference] The high-distance category defined by blocked or absent output (policy filters, missing languages) is arguably a different kind of phenomenon from mismatched content, and might be better modeled as a binary 'gate' rather than the far end of a continuous spectrum; the paper lumps both under one level.
  • [Editorial inference] The same lens could transfer outside education—journalism, healthcare, legal work—where global models meet local professional norms; the paper invites this but does not test it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper introduces 'cultural distance' as the gap between GenAI's default cultural repertoire and the situated demands of classroom teaching, and develops a three-level (low/mid/high) framework with six categories from 30 semi-structured interviews with K-12 teachers in South Africa, Taiwan, and the United States. The authors claim that teachers' alignment work clusters into two low-distance task types (communication, activity brainstorming), two mid-distance types (assessment generation, culturally relevant activity design), and two high-distance types (unsupported languages/traditions, policy restrictions). The paper offers design and policy implications aimed at reducing teachers' cultural labor.

Significance. If the taxonomy were robust, it would provide a useful middle-range vocabulary for HCI/CSCW research on AI alignment, invisible labor, and cross-cultural technology use, and it would be one of the few comparative qualitative studies of teacher-GenAI interaction across low- and high-resource settings. The study has real strengths: a clearly described semi-structured interview protocol, a demographic table, and rich, contextually grounded excerpts for each category. The cross-regional design with contexts chosen as a gradient is appropriate for the research question. However, the central classification is partly tautological, the effort-based levels are unstable within a task type, the sample truncates the outcome of interest, and the coding path is not auditable. These issues must be addressed before the framework can be credited as a transferable taxonomy.

major comments (4)
  1. [Section 3, Figure 1, Section 5 opening] The construct is operationalized as 'the varying amount of effort teachers must invest' (Figure 1 caption; Section 3), and Section 5 then 'identifies' three levels using that same effort criterion. The level labels are therefore a re-description of the coding variable rather than an independent empirical finding. The empirical content survives only in which tasks land at each level, but the claim 'we identified three levels of cultural distance' overstates the result. Please re-anchor the levels to independent features of the misalignment (e.g., linguistic representational gap, curricular mismatch, policy block) and treat effort as an outcome, or explicitly frame the taxonomy as an effort-based classification.
  2. [Section 5.2.1, M-1] The text reports that 'about half of the teachers' found assessment-question outputs satisfactory with little adjustment, while others required repeated prompting and revision; yet the entire task type is assigned to M-1 (mid-distance). This contradicts the level definition (mid = considerable effort) and shows the taxonomy cannot assign a stable effort signature to this task. Please report the distribution of teachers' reported effort per category, define subcategories conditional on teacher or student characteristics, or present M-1 as a context-dependent case rather than a stable level of distance.
  3. [Section 4.2 and Section 5.3] Recruitment required teachers who had already integrated GenAI into their practice (Section 4.2). The high-distance examples in Section 5.3 are therefore retrospective accounts of failure from continuing users; teachers who abandoned GenAI after such failures are excluded by design. This truncation of the outcome of interest means the 'unbridgeable' claim is supported only by survivors' recollections, not by disconfirming cases of abandonment. Please acknowledge this and discuss how dropout cases would affect the framework's generalizability.
  4. [Section 4.4 and Section 7] The coding path is not auditable. Section 4.4 describes reflexive thematic analysis but provides no codebook, no excerpt-to-code mapping beyond the one-line codes in Table 1, no inter-coder agreement or member checking, and no supplementary materials. For a taxonomy that asks readers to adopt six categories, the absence of an audit trail makes independent verification impossible. Please supply a coding appendix with category definitions, inclusion/exclusion criteria, and representative quotations, or make an anonymized codebook available.
minor comments (5)
  1. [Abstract and Section 1] 'offering teachers new ways for teaching practices' is ungrammatical; suggest 'new ways of supporting teaching practices' or similar.
  2. [Figure 1] The six categories are named in the caption but not defined in the figure; consider adding short definitions in the caption or a table, and reference the figure explicitly in Section 5.
  3. [Section 5.3.1] The sentence beginning 'While identifying similarities between minority cultures and AI's default cultural assumptions...' is orphaned and confusing; please revise or remove.
  4. [Table 1] Add a note explaining codes L-1, L-2, M-1, M-2, H-1, H-2 so the table is self-contained.
  5. [Section 7] The limitation paragraph on self-reported data could also acknowledge that the recruitment criteria may shape the reported effort distributions, pointing to the dropout issue raised in the major comments.

Circularity Check

1 steps flagged

Cultural-distance levels are defined by teacher effort, so the low/mid/high structure is partly a restatement of the coding criterion; the six task categories remain empirical.

specific steps
  1. self definitional [Figure 1 caption / Section 5 opening; Section 4.4; Section 5.2.1]
    "We identified three levels of cultural distance, each defined by the varying amount of effort teachers must invest when using GenAI to support their teaching practices, ranging from low to high. Within each level, we identified two distinct categories, for a total of six."

    The level construct is defined by the amount of effort teachers must invest, and the analysis clustered codes into themes reflecting 'different levels of effort and outcome' (Sec. 4.4). The finding that low-distance tasks need minimal adjustment, mid-distance tasks need considerable prompting/revision, and high-distance tasks cannot be bridged is therefore a restatement of the definitional criterion used to sort the data, not an independent property of the tasks. The paper's own data undercut a purely effort-based assignment: for assessment (labeled M-1), 'About half of the teachers reported that outputs were of satisfactory quality with little adjustment' (Sec. 5.2.1), so the same task type does not have a stable effort signature; the level label is an analyst judgment. The specific categ

full rationale

The central empirical contribution—six recurring categories of alignment work (L-1, L-2, M-1, M-2, H-1, H-2) with concrete examples from South Africa, Taiwan, and the U.S.—is not circular: the tasks assigned to each level were identified from interview data, and the paper explicitly presents the typology as open and exploratory (Sec. 1, Sec. 7). The three-level low/mid/high structure, however, is partly tautological because the levels are defined by the amount of effort teachers must invest and the coding procedure clustered themes by 'different levels of effort and outcome.' Thus the observation that low-distance tasks require little effort etc. restates the sorting rule. The paper's own data show within-task variability (e.g., assessment questions were satisfactory with little adjustment for about half of teachers, yet the task is labeled M-1), so the level assignments are analytic judgments rather than stable properties of the tasks. Self-citations to prior work by the authors (e.g., [23], [77], [80], [81]) are present but not load-bearing: the framework's categories are grounded in the 30 interviews, not derived from those citations. Because the category placements retain independent empirical content, the circularity is partial, not total; score 4.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 1 invented entities

The central claim rests on self-reported interviews and on the assumption that effort is a comparable measure of distance. There are no fitted numeric parameters. The main invented construct is 'cultural distance', which is empirically illustrated but not independently validated outside the paper.

axioms (4)
  • domain assumption Teachers' self-reported descriptions of GenAI use, effort, and outcomes are accurate reflections of classroom practice.
    Analysis in Section 4.4 and findings in Section 5 rely on retrospective interview accounts, with no classroom observation or usage logs to triangulate them, as the authors acknowledge in Section 7.
  • domain assumption The effort a teacher reports investing is a valid, cross-contextually comparable measure of the cultural distance between GenAI outputs and local demands.
    Section 3 and Figure 1 define low, mid, and high cultural distance by effort, assuming teachers with different prompt skills, persistence, and experience would report comparable effort for the same underlying gap.
  • domain assumption The three countries selected, South Africa, Taiwan, and the United States, form a gradient of cultural distance that supports generalizable conceptual claims.
    Section 4.1 justifies the contexts, but Section 6.1 and 6.4 generalize beyond these regions without probability sampling or tests of saturation.
  • domain assumption The six categories emerged inductively from reflexive thematic analysis rather than being imposed by the researchers' prior expectations.
    Section 4.4 describes reflexive thematic analysis, but the claim of emergence is not independently auditable because the codebook and coding decisions are not provided.
invented entities (1)
  • Cultural distance (as a new construct in GenAI education) no independent evidence
    purpose: Analytic lens to measure the gap between GenAI's default cultural repertoire and the situated demands of teaching practice, organized into low, mid, and high effort levels.
    The construct is grounded in 30 interviews but has no external falsifiable handle in the paper itself, and the established 'cultural distance' literature is not engaged.

reviewed 2026-08-04 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Bridging Cultural Distance Between Models Default and Local Classroom Demands: How Global Teachers Adopt GenAI to Support Everyday Teaching Practices." pith.science (2026). https://pith.science/paper/WV5KQTLK

@misc{pith2026250910780,
  author       = {Pith},
  title        = {Pith review of: Bridging Cultural Distance Between Models Default and Local Classroom Demands: How Global Teachers Adopt GenAI to Support Everyday Teaching Practices},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WV5KQTLK}},
  note         = {Machine review of arXiv:2509.10780}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Generative AI (GenAI) is rapidly entering K-12 classrooms, offering teachers new ways for teaching practices. Yet GenAI models are often trained on culturally uneven datasets, embedding a "default culture" that often misaligns with local classrooms. To understand how teachers navigate this gap, we defined the new concept Cultural Distance (the gap between GenAI's default cultural repertoire and the situated demands of teaching practice) and conducted in-depth interviews with 30 K-12 teachers, 10 each from South Africa, Taiwan, and the United States, who had integrated AI into their teaching practice. These teachers' experiences informed the development of our three-level cultural distance framework. This work contributes the concept and framework of cultural distance, six illustrative instances spanning in low, mid, high distance levels with teachers' experiences and strategies for addressing them. Empirically, we offer implications to help AI designers, policymakers, and educators create more equitable and culturally responsive GenAI tools for education.

Figures

Figures reproduced from arXiv: 2509.10780 by Hanqi Jane Li, Hong Shen, John Stamper, Phenyo Phemelo Moletsane, Qing Xiao, Ruiwei Xiao, Xinying Hou.

Figure 1
Figure 1. Figure 1: Teachers’ experiences with GenAI can be understood along a spectrum of [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

86 extracted references · 2 canonical work pages

  1. [1]

    Dhruv Agarwal, Mor Naaman, and Aditya Vashistha. 2025. AI suggestions homogenize writing toward western styles and diminish cultural nuances. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–21

    2025

  2. [2]

    Dakyeom Ahn and Hajin Lim. 2025. Exploring K-12 Physical Education Teach- ers’ Perspectives on Opportunities and Challenges of AI Integration through Ideation Workshops. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–16

    2025

  3. [3]

    Al Jazeera. 2024. What’s South Africa’s new school language law and why is it controversial. https://www.aljazeera.com/news/2024/9/18/whats-south-africas- new-school-language-law-and-why-is-it-controversial. Accessed July 16, 2025

    2024

  4. [4]

    2025.The Languages of South Africa

    Mary Alexander. 2025.The Languages of South Africa. https://southafrica- info.com/arts-culture/the-languages-of-south-africa/

    2025

  5. [5]

    Awad, Erika A

    Germine H. Awad, Erika A. Patall, Kadie R. Rackley, and Erin D. Reilly. 2015. Rec- ommendations for Culturally Sensitive Research Methods.Journal of Educational and Psychological Consultation25, 3 (2015), 283–303. https://doi.org/10.1080/ 10474412.2015.1046600

    arXiv 2015

  6. [6]

    Tara S Behrend, Eric N Wiebe, Jennifer E London, and Emily C Johnson. 2011. Cloud computing adoption and usage in community colleges.Behaviour & information technology30, 2 (2011), 231–240

    2011

  7. [7]

    Emily M Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. On the dangers of stochastic parrots: Can language models be too big?. InProceedings of the 2021 ACM conference on fairness, accountability, and transparency. 610–623

    2021

  8. [8]

    2025.Artificial Intelligence (AI) in K-8 Education: Understanding Teachers’ Perceptions and District Leader Readiness While Preparing for AI Adoption

    Anna Bernstein. 2025.Artificial Intelligence (AI) in K-8 Education: Understanding Teachers’ Perceptions and District Leader Readiness While Preparing for AI Adoption. Ph. D. Dissertation. Concordia University Chicago

    2025

  9. [9]

    Damian Blasi, Antonios Anastasopoulos, and Graham Neubig. 2021. Systematic inequalities in language technology performance across the world’s languages. arXiv preprint arXiv:2110.06733(2021)

    Pith/arXiv arXiv 2021

  10. [10]

    Virginia Braun and Victoria Clarke. 2019. Reflecting on reflexive thematic analysis. Qualitative research in sport, exercise and health11, 4 (2019), 589–597

    2019

  11. [11]

    BusinessTech. 2024. South Africa ranks as one of the biggest users of ChatGPT and AI globally. https://businesstech.co.za/news/technology/787980/south-africa- ranks-as-one-of-the-biggest-users-of-chatgpt-and-ai-globally/. Accessed Au- gust 27, 2025

    2024

  12. [12]

    Dan Calacci and Alex Pentland. 2022. Bargaining with the black-box: Designing and deploying worker-centric tools to audit algorithmic management.Proceedings of the ACM on Human-Computer Interaction6, CSCW2 (2022), 1–24

    2022

  13. [13]

    Margaret Chitiga, E Owusu-Sekyere, and N Tsoanamatsie. 2014. Income inequal- ity and limitations of the Gini index: The case of South Africa. (2014)

    2014

  14. [14]

    Irene-Angelica Chounta, Emanuele Bardone, Aet Raudsep, and Margus Pedaste

  15. [15]

    Clayton Cohn, Nicole Hutchins, Tuan Le, and Gautam Biswas. 2024. A chain- of-thought prompting approach with llms for evaluating students’ formative assessment responses in science. InProceedings of the AAAI conference on artificial intelligence, Vol. 38. 23182–23190

    2024

  16. [16]

    Jorge Cordero and Alison Cordero-Castillo. 2024. Exploring the Potential of Generative AI in Education: Opportunities, Challenges, and Best Practices for Classroom Integration. InWorld Congress in Computer Science, Computer Engi- neering & Applied Computing. Springer, 252–265

    2024

  17. [17]

    Council of Indigenous Peoples, Taiwan. 2024. Population and Demographics of Indigenous Peoples in Taiwan. Council of Indigenous Peoples, Executive Yuan. https://www.apc.gov.tw/portal/docList.html?CID=DC2E742A2C9F6E3A Accessed August 27, 2025

    2024

  18. [18]

    Mutlu Cukurova, Xin Miao, and Richard Brooker. 2023. Adoption of artificial intelligence in schools: Unveiling factors influencing teachers’ engagement. In International conference on artificial intelligence in education. Springer, 151–163

    2023

  19. [19]

    Mohamed Hassan Elnaem, Betul Okuyan, Naeem Mubarak, Abrar K Thabit, Merna Mahmoud AbouKhatwa, Diana Laila Ramatillah, AbdulMuminu Isah, Ali Azeez Al-Jumaili, and Nor Ilyani Mohamed Nazar. 2025. Students’ acceptance and use of generative AI in pharmacy education: international cross-sectional survey based on the extended unified theory of acceptance and us...

    2025

  20. [20]

    Ethnologue. 2025. South African Sign Language (SASL). https://www.ethnologue. com/language/sfs/. Accessed August 27, 2025

    2025

  21. [21]

    Fahim Faisal, Yinkai Wang, and Antonios Anastasopoulos. 2021. Dataset geogra- phy: Mapping language data to language users.arXiv preprint arXiv:2112.03497 (2021)

    Pith/arXiv arXiv 2021

  22. [22]

    Shabnam FakhrHosseini, Kathryn Chan, Chaiwoo Lee, Myounghoon Jeon, Heesuk Son, John Rudnik, and Joseph Coughlin. 2024. User adoption of in- telligent environments: A review of technology adoption models, challenges, and prospects.International Journal of Human–Computer Interaction40, 4 (2024), 986–998

    2024

  23. [23]

    Xianzhe Fan, Qing Xiao, Xuhui Zhou, Jiaxin Pei, Maarten Sap, Zhicong Lu, and Hong Shen. 2025. User-Driven Value Alignment: Understanding Users’ Perceptions and Strategies for Addressing Biased and Discriminatory Statements in AI Companions. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–19

    2025

  24. [24]

    Ozan Filiz, Mehmet Haldun Kaya, and Tufan Adiguzel. 2025. Teachers and AI: Understanding the factors influencing AI integration in K-12 education.Education and Information Technologies(2025), 1–37

    2025

  25. [25]

    Iason Gabriel. 2020. Artificial intelligence, values, and alignment.Minds and machines30, 3 (2020), 411–437

    2020

  26. [26]

    Luise Ge, Daniel Halpern, Evi Micha, Ariel D Procaccia, Itai Shapira, Yevgeniy Vorobeychik, and Junlin Wu. 2024. Axioms for ai alignment from human feedback. Advances in Neural Information Processing Systems37 (2024), 80439–80465

    2024

  27. [27]

    Google DeepMind. 2023. Gemini. https://deepmind.google/technologies/gemini/. Accessed May 6, 2025

    2023

  28. [28]

    Grand View Research. 2025. AI In Education Market Size, Share & Trends Analysis Report By Component, By Deployment, By Technology (NLP, ML), By Applica- tion (Intelligent Tutoring System, Learning Platform & Virtual Facilitators), By End-use, By Region, And Segment Forecasts, 2025 – 2030. Grand View Research In- dustry Report. https://www.grandviewresearc...

    2025

  29. [29]

    Dylan Hadfield-Menell and Gillian K Hadfield. 2019. Incomplete contracting and AI alignment. InProceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society. 417–422

    2019

  30. [30]

    Karen Hao. 2022. Artificial intelligence is creating a new colonial world order. MIT Technology Review(2022)

    2022

  31. [31]

    Ellie Harmon and M Six Silberman. 2019. Rating working conditions on digital labor platforms.Computer Supported Cooperative Work (CSCW)28, 5 (2019), 911–960

    2019

  32. [32]

    Don’t Forget the Teachers

    Emma Harvey, Allison Koenecke, and Rene F Kizilcec. 2025. " Don’t Forget the Teachers": Towards an Educator-Centered Understanding of Harms from Large Language Models in Education. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–19

    2025

  33. [33]

    Michael A Hedderich, Natalie N Bazarova, Wenting Zou, Ryun Shim, Xinda Ma, and Qian Yang. 2024. A piece of theatre: Investigating how teachers design LLM chatbots to assist adolescent cyberbullying education. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems. 1–17

    2024

  34. [34]

    Yoko Hirata and Yoshihiro Hirata. 2025. How Do Students’ Social and Educa- tional Norms and Practices Affect GenAI Chatbot-Facilitated Conversations?SN Computer Science6, 4 (2025), 1–11

    2025

  35. [35]

    Kristina Höök. 2000. Steps to take before intelligent user interfaces become real. Interacting with computers12, 4 (2000), 409–426

    2000

  36. [36]

    Xinying Hou, Carol Forsyth, Jessica Andrews-Todd, James Rice, Zhiqiang Cai, Yang Jiang, Diego Zapata-Rivera, and Art Graesser. 2025. An LLM-Enhanced Multi-agent Architecture for Conversation-Based Assessment. InInternational Conference on Artificial Intelligence in Education. Springer, 119–134

    2025

  37. [37]

    Sarah Howie, Elsie Venter, and Surette Van Staden. 2008. The effect of multilingual policies on performance and progression in reading literacy in South African primary schools.Educational Research and Evaluation14, 6 (2008), 551–560

    2008

  38. [38]

    Pratik Joshi, Sebastin Santy, Amar Budhiraja, Kalika Bali, and Monojit Choudhury

  39. [39]

    Kostas Karpouzis, Dimitris Pantazatos, Joanna Taouki, and Kalliopi Meli. 2024. Tailoring education with GenAI: a new horizon in lesson planning. In2024 IEEE Global Engineering Education Conference (EDUCON). IEEE, 1–10

    2024

  40. [40]

    Majeed Kazemitabaar, Xinying Hou, Austin Henley, Barbara Jane Ericson, David Weintrop, and Tovi Grossman. 2023. How novices use LLM-based code generators to solve CS1 coding tasks in a self-paced learning environment. InProceedings How Global Teachers Use GenAI to Support Everyday Teaching Practices Conference’17, July 2017, Washington, DC, USA of the 23r...

    2023

  41. [41]

    Khan Academy. 2023. Khanmigo: AI Tutor by Khan Academy. https://www. khanacademy.org/khan-labs. Accessed May 6, 2025

    2023

  42. [42]

    Maria Klar. 2025. Using ChatGPT is easy, using it effectively is tough? A mixed methods study on K-12 students’ perceptions, interaction patterns, and support for learning with generative AI chatbots.Smart Learning Environments12, 1 (2025), 32

    2025

  43. [43]

    Yuxuan Li, Hirokazu Shirado, and Sauvik Das. 2025. Actions speak louder than words: Agent decisions reveal implicit biases in language models. InProceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency. 3303– 3325

    2025

  44. [44]

    Ally Limke, Saminur Islam, Bahare Riahi, Xiaoyi Tian, Marnie Hill, Veronica Catété, and Tiffany Barnes. 2025. What Does It Take to Support Problem Solving in Programming Classrooms? A New Framework from the K-12 Teacher Perspec- tive. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems Extended Abstracts (CHI EA ’25). Associatio...

    arXiv 2025

  45. [45]

    Xin Lin, Xiaonan Han, and Junqiao Qiu. 2025. Generative AI in Special Educa- tion: Teachers’ Insights on Instructional Enrichment vs. Accommodations. In Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems. 1–6

    2025

  46. [46]

    Zhaoming Liu. 2025. Cultural bias in large language models: A comprehensive analysis and mitigation strategies.Journal of Transcultural Communication3, 2 (2025), 224–244

    2025

  47. [47]

    Jackson G Lu, Lesley Luyang Song, and Lu Doris Zhang. 2025. Cultural tendencies in generative AI.Nature Human Behaviour(2025), 1–10

    2025

  48. [48]

    Hanjia Lyu, Jiebo Luo, Jian Kang, and Allison Koenecke. 2025. Characterizing Bias: Benchmarking Large Language Models in Simplified versus Traditional Chinese. InProceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency. 2815–2846

    2025

  49. [49]

    2024.The impact of artificial intelligence tools on bilingual students in us education: a study on academic language-learning, cultural sensitivity and inclusiveness

    Baoyi Ma. 2024.The impact of artificial intelligence tools on bilingual students in us education: a study on academic language-learning, cultural sensitivity and inclusiveness. University of Washington

    2024

  50. [50]

    Shuaiyao Ma and Lei Lei. 2024. The factors influencing teacher education students’ willingness to adopt artificial intelligence technology for information-based teaching.Asia Pacific Journal of Education44, 1 (2024), 94–111

    2024

  51. [51]

    MagicSchool AI. 2023. MagicSchool AI. https://www.magicschool.ai/. Accessed May 6, 2025

    2023

  52. [52]

    2025.Artificial Intelligence Index Report 2025

    Nestor Maslej, Loredana Fattorini, Raymond Perrault, Yolanda Gil, Vanessa Parli, Njenga Kariuki, Emily Capstick, Anka Reuel, Erik Brynjolfsson, John Etchemendy, Katrina Ligett, Terah Lyons, James Manyika, Juan Carlos Niebles, Yoav Shoham, Russell Wald, et al. 2025.Artificial Intelligence Index Report 2025. https://hai. stanford.edu/ai-index/2025-ai-index-report

    2025

  53. [53]

    Nestor Maslej, Loredana Fattorini, Raymond Perrault, Yolanda Gil, Vanessa Parli, Njenga Kariuki, Emily Capstick, Anka Reuel, Erik Brynjolfsson, John Etchemendy, Katrina Ligett, Terah Lyons, James Manyika, Juan Carlos Niebles, Yoav Shoham, Russell Wald, et al. 2025. Artificial Intelligence Index Report 2025. Stanford Institute for Human-Centered Artificial...

    2025

  54. [54]

    Tayab D Memon and Paul Kwan. 2025. A Collaborative Model for Integrating Teacher and GenAI into Future Education.TechTrends(2025), 1–15

    2025

  55. [55]

    1949.On sociological theories of the middle range [1949]

    Robert King Merton. 1949.On sociological theories of the middle range [1949]. na

    1949

  56. [56]

    I Would Never Trust Anything Western

    Manas Mhasakar, Rachel Baker-Ramos, Benjamin Carter, Evyn-Bree Helekahi- Kaiwi, and Josiah Hester. 2025. " I Would Never Trust Anything Western": Kumu (Educator) Perspectives on Use of LLMs for Culturally Revitalizing CS Education in Hawaiian Schools. InProceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems. 1–10

    2025

  57. [57]

    I Would Never Trust Anything Western

    Manas Mhasakar, Rachel Baker-Ramos, Benjamin Carter, Evyn-Bree Helekahi- Kaiwi, and Josiah Hester. 2025. “I Would Never Trust Anything Western”: Kumu (Educator) Perspectives on Use of LLMs for Culturally Revitalizing CS Education in Hawaiian Schools. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems Extended Abstracts (CHI EA ...

    arXiv 2025

  58. [58]

    Nicolae Nistor, Aytaç Göğüş, and Thomas Lerche. 2013. Educational technol- ogy acceptance across national and professional cultures: a European study. Educational Technology Research and Development61, 4 (2013), 733–749

    2013

  59. [59]

    OpenAI. 2022. ChatGPT. https://chat.openai.com/. Accessed May 6, 2025

    2022

  60. [60]

    Maximino Plata. 2009. Cultural sensitivity: The basis for culturally relevant teaching.Tep21, 2 (2009), 181

    2009

  61. [61]

    Associated Press. 2025. AI tools helping teachers reclaim valuable time. Gallup–Walton Family Foundation poll. 6 in 10 teachers used AI in 2024–25, weekly users save 5.9 hours/week

    2025

  62. [62]

    Katharina Reinecke and Abraham Bernstein. 2011. Improving performance, perceived usability, and aesthetics with culturally adaptive user interfaces.ACM Trans. Comput.-Hum. Interact.18, 2, Article 8 (July 2011), 29 pages. https://doi. org/10.1145/1970378.1970382

    arXiv 2011

  63. [63]

    Katharina Reinecke and Abraham Bernstein. 2013. Knowing what a user likes: A design science approach to interfaces that automatically adapt to culture.Mis Quarterly(2013), 427–453

    2013

  64. [64]

    2023.Cultures in human-computer interaction

    Sergio Sayago. 2023.Cultures in human-computer interaction. Springer

    2023

  65. [65]

    2008.Class, race, and inequality in South Africa

    Jeremy Seekings and Nicoli Nattrass. 2008.Class, race, and inequality in South Africa. Yale University Press

    2008

  66. [66]

    Hua Shen, Tiffany Knearem, Reshmi Ghosh, Kenan Alkiek, Kundan Krishna, Yachuan Liu, Ziqiao Ma, Savvas Petridis, Yi-Hao Peng, Li Qiwei, et al . 2024. Towards bidirectional human-ai alignment: A systematic review for clarifications, framework, and future directions.arXiv preprint arXiv:2406.09264(2024)

    arXiv 2024

  67. [67]

    Hua Shen, Tiffany Knearem, Reshmi Ghosh, Michael Xieyang Liu, Andrés Monroy-Hernández, Tongshuang Wu, Diyi Yang, Yun Huang, Tanushree Mitra, Yang Li, et al. 2025. Bidirectional Human-AI Alignment: Emerging Challenges and Opportunities. InProceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems. 1–6

    2025

  68. [68]

    Hua Shen, Ziqiao Ma, Reshmi Ghosh, Tiffany Knearem, Michael Xieyang Liu, Tongshuang Wu, Andrés Monroy-Hernández, Diyi Yang, Antoine Bosselut, Furong Huang, et al. [n. d.]. ICLR 2025 Workshop on Bidirectional Human-AI Alignment. InICLR 2025 Workshop Proposals

    2025

  69. [69]

    Smith, Julia Hudnut-Beumler, and Seth J

    Ashley E. Smith, Julia Hudnut-Beumler, and Seth J. Scholer. 2017. Can Discipline Education be Culturally Sensitive?Maternal and Child Health Journal21 (2017), 177–186. https://doi.org/10.1007/s10995-016-2107-9

    work page doi:10.1007/s10995-016-2107-9 2017

  70. [70]

    Nic Spaull and Jonathan D Jansen. 2019. South African schooling: The enigma of inequality.Switzerlan: Springer Nature(2019)

    2019

  71. [71]

    Statista. 2025. Countries with the Highest Income Inequality 2025 (Gini In- dex). https://www.statista.com/statistics/264627/ranking-of-the-20-countries- with-the-biggest-inequality-in-income-distribution/. Accessed July 16, 2025

    2025

  72. [72]

    Evan T Straub. 2009. Understanding technology adoption: Theory and future directions for informal learning.Review of educational research79, 2 (2009), 625–649

    2009

  73. [73]

    Yan Tao, Olga Viberg, Ryan S Baker, and René F Kizilcec. 2024. Cultural bias and cultural alignment of large language models.PNAS nexus3, 9 (2024), pgae346

    2024

  74. [74]

    Margaret Yun-Pu Tu. 2025. AI, algorithms and Indigenous agency in Taiwan. East Asia Forum, 25 April 2025. https://eastasiaforum.org/2025/04/25/ai-algorithms- and-indigenous-agency-in-taiwan/ Accessed August 27, 2025

    2025

  75. [75]

    Kizilcec

    Olga Viberg, Mutlu Cukurova, Yael Feldman-Maggor, Giora Alexandron, Shizuka Shirai, Susumu Kanemune, Barbara Wasson, Cathrine Tømte, Daniel Spikol, Marcelo Milrad, Raquel Coelho, and René F. Kizilcec. 2024. What Explains Teachers’ Trust of AI in Education across Six Countries?International Journal of Artificial Intelligence in Education(2024). https://doi...

    work page doi:10.1007/s40593-024- 2024

  76. [76]

    Deliang Wang, Dapeng Shan, Ran Ju, Ben Kao, Chenwei Zhang, and Gaowei Chen

  77. [77]

    Jiayi Wang, Ruiwei Xiao, Xinying Hou, Hanqi Li, Ying Jui Tseng, John Stamper, and Kenneth Koedinger. 2025. LLMs to Support K–12 Teachers in Culturally Relevant Pedagogy: An AI Literacy Example. InInternational Conference on Artificial Intelligence in Education. Springer, 152–160

    2025

  78. [78]

    Karen Woodruff, James Hutson, and Kathryn Arnone. 2023. Perceptions and barriers to adopting artificial intelligence in K-12 education: A survey of educators in fifty states. (2023)

    2023

  79. [79]

    Di Wu, Xinyan Zhang, Kaili Wang, Longkai Wu, and Wei Yang. 2025. A multi-level factors model affecting teachers’ behavioral intention in AI-enabled education ecosystem.Educational Technology Research and Development73, 1 (2025), 135– 167

    2025

  80. [80]

    It Might be Technically Impressive, But It’s Practically Useless to us

    Qing Xiao, Xianzhe Fan, Felix Marvin Simon, Bingbing Zhang, and Motahhare Eslami. 2025. " It Might be Technically Impressive, But It’s Practically Useless to us": Motivations, Practices, Challenges, and Opportunities for Cross-Functional Collaboration around AI within the News Industry. InProceedings of the 2025 CHI Conference on Human Factors in Computin...

    2025

Showing first 80 references.

This paper was first reviewed by deepseek-v4-flash on August 4, 2026.