pith. sign in

arxiv: 2604.04139 · v1 · submitted 2026-04-05 · 💻 cs.HC

Teacher Professional Development on WhatsApp and LLMs: Early Lessons from Cameroon

Vikram Kamath Cannanure , Bruno Yinkfu , Douglas Bryan , Mati Amin , Ingmar Weber This is my paper

Pith reviewed 2026-05-13 16:41 UTC · model grok-4.3

classification 💻 cs.HC
keywords teacher professional developmentWhatsApp chatbotLLM-supported contentusability evaluationlow-resource settingsCameroonmobile learningAI accessibility
0
0 comments X

The pith

A WhatsApp chatbot with LLM content achieves higher usability ratings than online forms for teacher training in Cameroon.

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

This paper tests whether AI for teacher professional development can reach low-resource settings by running on everyday mobile messaging instead of web platforms. In a pilot with 47 primary school teachers, the WhatsApp chatbot received better ratings for usability and overall experience than a standard online form while learnability stayed the same. The gains came from teachers already knowing the app, needing fewer steps to interact, and receiving content in short modules generated by language models. Limits appeared around unreliable connections, prepaid data costs, and the need to handle both English and French. The work shows how fitting AI tools to existing habits can support ongoing teacher growth in places where institutional websites are impractical.

Core claim

In a mixed-methods study with 47 primary school teachers in Cameroon, a WhatsApp-based chatbot incorporating LLM-supported content for professional development was evaluated against an online form. The chatbot received higher ratings for perceived usability and overall experience, while learnability was similar. These advantages arose from the familiarity of the WhatsApp platform, lower interaction demands, and the modular organization of the AI-generated materials. Limitations related to network access, data expenses, and bilingual needs were noted, leading to proposed design principles for culturally appropriate and reflective AI interactions.

What carries the argument

WhatsApp-based LLM chatbot delivering modular professional development content, which carries the argument by reducing access barriers through everyday platform use and structured short interactions.

Load-bearing premise

That the higher usability ratings are caused mainly by platform familiarity and low interaction overhead rather than differences in the specific training content or the teachers who participated.

What would settle it

A larger randomized trial that holds content quality constant across conditions and measures usability again; if the chatbot no longer shows an advantage, the central claim would not hold.

Figures

Figures reproduced from arXiv: 2604.04139 by Bruno Yinkfu, Douglas Bryan, Ingmar Weber, Mati Amin, Vikram Kamath Cannanure.

Figure 1
Figure 1. Figure 1: In-person teacher professional development workshop conducted as part of the study. Teachers participated in facilitated group discussions and reflective activities that complemented the WhatsApp-based chatbot learning modules. Partner Organization and Training Program The study was conducted in col￾laboration with an international organization that provides in-service TPD for primary school teachers acros… view at source ↗
Figure 2
Figure 2. Figure 2: Example WhatsApp chatbot lesson flow used during the teacher professional development pilot. Screenshots show (left) course selection and multimedia lesson con￾tent; (middle) a reflection prompt embedded in the chat interaction; and (right) lesson completion and navigation to subsequent modules. brief school-based group discussions were held across six schools (two repre￾sentatives per school, English/Fren… view at source ↗
Figure 3
Figure 3. Figure 3: Instructor-facing dashboard for managing WhatsApp-based professional de￾velopment content. Left: course analytics and lesson structure. Right: item creation interface with real-time WhatsApp preview. 30. World Bank: From chalkboards to chatbots: Evaluating the impact of generative ai on learning outcomes in nigeria (2024), https://openknowledge.worldbank.org/ entities/publication/15e1ff08-15ae-4f7a-b2a8-d1… view at source ↗
read the original abstract

AI in education is commonly delivered through web-based systems such as online forms and institutional platforms. However, these approaches can exclude teachers in low-resource contexts, where everyday mobile platforms like WhatsApp serve as primary digital infrastructure. To address this gap, we present a field pilot in Cameroon that deploys a WhatsApp-based chatbot with LLM-supported content for teacher professional development (TPD), compared with an online form baseline. The system was evaluated through a mixed-methods study with 47 primary school teachers, integrating quantitative measures with qualitative insights from interviews and participant feedback. Results show that the chatbot was rated higher in perceived usability and overall experience, while learnability remained comparable. These improvements were driven by platform familiarity, low interaction overhead, and the modular structure of LLM-supported content, but were constrained by connectivity limitations, prepaid data costs, and multilingual needs (English/French). Building on these findings, we outline design directions for multilingual, culturally grounded interaction and for supporting prompting and reflection in AI use. More broadly, this work points to Thoughtful AI that supports reflection, relevance, and sustained professional growth.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper presents a field pilot in Cameroon deploying a WhatsApp-based chatbot with LLM-supported content for teacher professional development (TPD), compared to an online form baseline. A mixed-methods evaluation with 47 primary school teachers finds higher perceived usability and overall experience ratings for the chatbot (learnability comparable), attributing gains to platform familiarity, low interaction overhead, and modular LLM content while noting constraints from connectivity, data costs, and multilingual needs; it outlines design directions for culturally grounded AI interactions supporting reflection.

Significance. If the usability findings hold after addressing design limitations, the work offers timely evidence on leveraging ubiquitous mobile platforms like WhatsApp to make AI-enhanced TPD more accessible in low-resource settings, contributing concrete early lessons on interaction overhead, modularity, and multilingual support that could inform inclusive HCI design for education in similar contexts.

major comments (3)
  1. [Methods] Methods: The manuscript does not describe randomization of the 47 teachers to WhatsApp vs. online-form conditions or exact matching of TPD content across arms; without these, the claim that usability gains are 'driven by' platform familiarity (vs. content differences or self-selection) cannot be isolated.
  2. [Results] Results: No statistical tests, confidence intervals, or effect sizes are reported for the quantitative usability and experience ratings despite N=47; this leaves the magnitude and reliability of the chatbot advantage unclear and weakens causal attribution.
  3. [Discussion] Discussion: Qualitative themes supporting the 'low overhead' and 'modular LLM content' explanations are presented without controls for teacher demographics, prior tech experience, or content quality, so alternative explanations (e.g., novelty or selection bias) remain unruled out.
minor comments (2)
  1. [Abstract] Abstract and §1: The term 'Thoughtful AI' is introduced without an explicit definition or operationalization; a brief parenthetical or footnote would improve clarity.
  2. [Results] Figure captions and tables: Ensure all quantitative rating scales (e.g., SUS or custom Likert items) and exact question wording are reproduced so readers can assess comparability across conditions.

Simulated Author's Rebuttal

3 responses · 0 unresolved

Thank you for the constructive feedback on our manuscript. We value the referee's insights on strengthening the methodological rigor and presentation of results. We have carefully considered each comment and provide point-by-point responses below. Where appropriate, we will revise the manuscript to address the concerns.

read point-by-point responses

  1. Referee: [Methods] Methods: The manuscript does not describe randomization of the 47 teachers to WhatsApp vs. online-form conditions or exact matching of TPD content across arms; without these, the claim that usability gains are 'driven by' platform familiarity (vs. content differences or self-selection) cannot be isolated.

    Authors: We appreciate this observation. Our study was designed as an early field pilot rather than a controlled experiment, with teachers assigned to conditions based on school-level logistics and availability to minimize disruption in the Cameroonian primary school context. The TPD content was developed to be equivalent across arms, drawing from the same curriculum modules, though exact matching was adapted for platform affordances. We will revise the Methods section to explicitly describe the assignment process, content equivalence efforts, and acknowledge the lack of randomization as a limitation. We will also adjust language from 'driven by' to 'associated with' to avoid implying causality. revision: yes

  2. Referee: [Results] Results: No statistical tests, confidence intervals, or effect sizes are reported for the quantitative usability and experience ratings despite N=47; this leaves the magnitude and reliability of the chatbot advantage unclear and weakens causal attribution.

    Authors: We agree that reporting inferential statistics would enhance the quantitative findings. Although the primary focus was on mixed-methods insights from this pilot, we have access to the raw usability scores and will include Mann-Whitney U tests (given non-normal distributions), 95% confidence intervals, and Cohen's d effect sizes in the revised Results section. This will provide a clearer picture of the effect magnitude while noting the exploratory nature of the analysis. revision: yes

  3. Referee: [Discussion] Discussion: Qualitative themes supporting the 'low overhead' and 'modular LLM content' explanations are presented without controls for teacher demographics, prior tech experience, or content quality, so alternative explanations (e.g., novelty or selection bias) remain unruled out.

    Authors: The qualitative themes emerged from thematic analysis of interviews and feedback, and we recognize the value of addressing potential confounds. In the revised Discussion, we will expand the limitations paragraph to explicitly discuss novelty effects, selection bias, and the absence of controls for demographics and prior experience. We will also report any available demographic data and suggest directions for future controlled studies. As this was an early pilot study, implementing full controls was beyond the current scope. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical claims rest on direct study data

full rationale

The paper reports a mixed-methods field pilot with 47 teachers, using quantitative usability ratings and qualitative interviews to compare a WhatsApp LLM chatbot against an online-form baseline. All central claims (higher perceived usability and experience for the chatbot, driven by platform familiarity and low overhead) are presented as direct outputs of the collected data and participant feedback. No equations, fitted parameters, predictions, or self-citation chains appear; the attribution narrative is interpretive but grounded in the study's own observations rather than reducing to prior inputs by construction. This is a standard empirical evaluation with no load-bearing derivations.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on empirical observations from a small-scale pilot; no free parameters or invented entities, but relies on assumptions about the representativeness of the sample and attribution of improvements.

axioms (1)
  • domain assumption The mixed-methods evaluation with 47 teachers provides valid insights into usability and experience.
    Assumed in the study design without detailed justification in abstract.

pith-pipeline@v0.9.0 · 5503 in / 1185 out tokens · 60726 ms · 2026-05-13T16:41:28.366153+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

  1. [1]

    Journal of Multilingual and Multicultural Development27(2), 96–112 (2006)

    Anchimbe, E.: Functional seclusion and the future of indigenous languages in cameroon. Journal of Multilingual and Multicultural Development27(2), 96–112 (2006)

    work page 2006

  2. [2]

    Knowledge and Skills for development, Brighton1(1), 1–52 (2004)

    Bennell, P.: Teacher motivation and incentives in sub-saharan africa and asia. Knowledge and Skills for development, Brighton1(1), 1–52 (2004)

    work page 2004

  3. [3]

    In: Proceedings of the 2020 International Conference on Information and Communication Technologies and Development

    Cannanure, V.K., Brown, T.X., Ogan, A.: Dia: A human ai hybrid conversa- tional assistant for developing contexts. In: Proceedings of the 2020 International Conference on Information and Communication Technologies and Development. ICTD2020, Association for Computing Machinery, New York, NY, USA (2020). https://doi.org/10.1145/3392561.3397577

    work page doi:10.1145/3392561.3397577 2020

  4. [4]

    Cannanure, V.K., Ngoon, T., Wolf, S., Jasinska, K., Brown, T., Ogan, A.: A design- based research approach to bridge teacher aspirations and goal-setting. ACM J. Comput. Sustain. Soc.3(4) (Nov 2025),https://doi.org/10.1145/3736648

    work page doi:10.1145/3736648 2025

  5. [5]

    Cannanure, V.K., Ngoon, T., Wolf, S., Jasińska, K., Brown, T.X., Ogan, A.: Un- derstanding the longitudinal impact of a chatbot to facilitate a virtual community of practice for teachers in rural côte d’ivoire. ACM J. Comput. Sustain. Soc.2(3) (Sep 2024).https://doi.org/10.1145/3675762

    work page doi:10.1145/3675762 2024

  6. [6]

    In: Proceedings of the 3rd ACM SIGCAS Conference on Computing and Sustainable Societies

    Cannanure, V.K., Niweteto, J.S., Adji, Y.T., Hermann, A.Y., Jasinska, K.K., Brown, T.X., Ogan, A.: I’m fine where i am, but i want to do more: Explor- ing teacher aspirations in rural côte d’ivoire. In: Proceedings of the 3rd ACM SIGCAS Conference on Computing and Sustainable Societies. p. 1–12. COM- PASS ’20, Association for Computing Machinery, New York...

    work page doi:10.1145/3378393.3403653 2020

  7. [7]

    We dream of climbing the ladder; to get there, we have to do our job better

    Cannanure, V.K., Ávila Uribe, E., Ngoon, T., Adji, Y., Wolf, S., Jasińska, K., Brown, T., Ogan, A.: “We dream of climbing the ladder; to get there, we have to do our job better”: Designing for Teacher Aspirations in rural Côte d’Ivoire. In: ACM SIGCAS/SIGCHI Conference on Computing and Sustainable Societies (COMPASS). pp. 122–138. ACM, Seattle WA USA (Jun...

    work page doi:10.1145/3530190.3534794 2022

  8. [8]

    In: Hesse-Biber, S., (eds.), P.L.(eds.)Handbookofemergentmethods,pp.155–170.TheGuilfordPress(2008)

    Charmaz, K.: Grounded theory as an emergent method. In: Hesse-Biber, S., (eds.), P.L.(eds.)Handbookofemergentmethods,pp.155–170.TheGuilfordPress(2008)

    work page 2008

  9. [9]

    DBER Speaker Series (2013)

    Creswell, J.W.: Steps in conducting a scholarly mixed methods study. DBER Speaker Series (2013)

    work page 2013

  10. [10]

    Early Education and Development23(5), 680– 699 (2012)

    Denham, S., Brown, C.: Social-emotional learning in early childhood: What we know and where to go from here. Early Education and Development23(5), 680– 699 (2012)

    work page 2012

  11. [11]

    Linguistik Online18, 19–33 (2004)

    Echu, G.: The language question in cameroon. Linguistik Online18, 19–33 (2004)

    work page 2004

  12. [12]

    Love and Logic Press (1995)

    Fay, J., Funk, D.: Teaching with Love and Logic. Love and Logic Press (1995)

    work page 1995

  13. [13]

    Conference on Human Factors in Computing Systems - Proceedings pp

    Grudin, J., Jacques, R.: Chatbots, humbots, and the quest for artificial general intelligence. Conference on Human Factors in Computing Systems - Proceedings pp. 1–11 (2019).https://doi.org/10.1145/3290605.3300439

    work page doi:10.1145/3290605.3300439 2019

  14. [14]

    GSMA: The mobile economy sub-saharan africa 2023 (2023) Teacher Professional Development on WhatsApp and LLMs... 11

    work page 2023

  15. [15]

    Oxford University Press (2009)

    Hirsh-Pasek, K., Golinkoff, R., Berk, L.: A mandate for playful learning in preschool: Presenting the evidence. Oxford University Press (2009)

    work page 2009

  16. [16]

    International Telecommunication Union: Measuring digital development: Facts and figures (2023)

    work page 2023

  17. [17]

    Teaching and Teacher Education168, 105219 (2025),https:// www.sciencedirect.com/science/article/pii/S0742051X25002963

    Li, K., Wang, P., Chen, G.: How can ai be integrated into teacher professional development programs? a systematic review based on an adapted technology-based learning model. Teaching and Teacher Education168, 105219 (2025),https:// www.sciencedirect.com/science/article/pii/S0742051X25002963

    work page 2025

  18. [18]

    Current Issues in Comparative Education20(Jan 2018)

    Mary Mendenhall, Makala Skinner, Sophia Collas, Sarah French: Expanding Teacher Support through Mobile Mentoring in Kakuma Refugee Camp: Bene- fits and Challenges. Current Issues in Comparative Education20(Jan 2018). https://doi.org/10.52214/cice.v20i.11549

    work page doi:10.52214/cice.v20i.11549 2018

  19. [19]

    Computers & Education132, 139–153 (2019)

    Motteram, G., Dawson, S.: Resilient professional learning with whatsapp teacher communities. Computers & Education132, 139–153 (2019)

    work page 2019

  20. [20]

    Education and Infor- mation Technologies25(6), 5731–5751 (Nov 2020).https://doi.org/10.1007/ s10639-020-10233-0

    Motteram, G., Dawson, S., Al-Masri, N.: WhatsApp supported language teacher development: A case study in the Zataari refugee camp. Education and Infor- mation Technologies25(6), 5731–5751 (Nov 2020).https://doi.org/10.1007/ s10639-020-10233-0

    work page 2020

  21. [21]

    Accessed: 2026-02-16

    Mozilla Foundation: Lessons from building for kinyarwanda on com- mon voice (2023),https://www.mozillafoundation.org/en/blog/ lessons-from-building-for-kinyarwanda-on-common-voice/, mozilla Founda- tion Blog. Accessed: 2026-02-16

    work page 2023

  22. [22]

    Rome: FAO p

    Mulkeen, A.: Teachers for rural schools: A challenge for africa. Rome: FAO p. 8 (2005)

    work page 2005

  23. [23]

    Computers & Education173, 104269 (2021)

    Nelimarkka, M., Leinonen, T., Durall, E., Dean, P.: Facebook is not a silver bullet for teachers’ professional development: Anatomy of an eight-year-old social-media community. Computers & Education173, 104269 (2021)

    work page 2021

  24. [24]

    In: Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems

    Poon, A., Giroux, S., Eloundou-Enyegue, P., Guimbretiere, F., Dell, N.: Engaging highschoolstudentsincameroonwithexampracticequizzesviasmsandwhatsapp. In: Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems. p. 1–13. CHI ’19, Association for Computing Machinery, New York, NY, USA (2019).https://doi.org/10.1145/3290605.3300712

    work page doi:10.1145/3290605.3300712 2019

  25. [25]

    Computers and Education: Ar- tificial Intelligence8, 100355 (2025),https://www.sciencedirect.com/science/ article/pii/S2666920X24001589

    Tan, X., Cheng, G., Ling, M.H.: Artificial intelligence in teaching and teacher professional development: A systematic review. Computers and Education: Ar- tificial Intelligence8, 100355 (2025),https://www.sciencedirect.com/science/ article/pii/S2666920X24001589

    work page 2025

  26. [26]

    The Stoplight Approach International Association: Stoplight starts with me: Im- pact overview.https://www.thestoplightapproach.org/impact-2(2026), ac- cessed: 2026-02-11

    work page 2026

  27. [27]

    Proceedings of the ACM on Human-Computer Interaction3(CSCW), 1–21 (Nov 2019).https: //doi.org/10.1145/3359322

    Varanasi, R.A., Kizilcec, R.F., Dell, N.: How Teachers in India Reconfigure their Work Practices around a Teacher-Oriented Technology Intervention. Proceedings of the ACM on Human-Computer Interaction3(CSCW), 1–21 (Nov 2019).https: //doi.org/10.1145/3359322

    work page doi:10.1145/3359322 2019

  28. [28]

    In: Proceed- ings of the 2021 CHI Conference on Human Factors in Computing Systems

    Varanasi, R.A., Vashistha, A., Dell, N.: Tag a Teacher: A Qualitative Analysis of WhatsApp-Based Teacher Networks in Low-Income Indian Schools. In: Proceed- ings of the 2021 CHI Conference on Human Factors in Computing Systems. pp. 1–16. ACM, Yokohama Japan (May 2021).https://doi.org/10.1145/3411764. 3445221

    work page doi:10.1145/3411764 2021

  29. [29]

    Mind, Brain, and Education7(2), 104–112 (2013) 12 V

    Weisberg, D., Hirsh-Pasek, K., Golinkoff, R.: Guided play: Where curricular goals meet a playful pedagogy. Mind, Brain, and Education7(2), 104–112 (2013) 12 V. Cannanure et al. (a) Course dashboard showing enrollments, active learners, and lesson modules. (b) Authoring interface for creating chatbot learning items and previewing WhatsApp delivery. Fig. 3....

    work page 2013

  30. [30]

    Teaching language and WhatsApp use were self-reported in the post-training survey

    World Bank: From chalkboards to chatbots: Evaluating the impact of generative ai on learning outcomes in nigeria (2024),https://openknowledge.worldbank.org/ entities/publication/15e1ff08-15ae-4f7a-b2a8-d146e6c113ee, world Bank Open Knowledge Repository 4 Appendix T able 3.Participant demographics and technology use (n= 14). Teaching language and WhatsApp ...

    work page 2024