pith. sign in

arxiv: 2605.15376 · v1 · pith:GCMG72XVnew · submitted 2026-05-14 · 💻 cs.CL · cs.CY

Adesua: Development and Feasibility Study of an AI WhatsApp Bot for Science Learning in West Africa

George Boateng , Evans Atompoya , Philemon Badu , Samuel John , Samuel Ansah , Patrick Agyeman-Budu , Victor Wumbor-Apin Kumbol This is my paper

Pith reviewed 2026-05-19 15:18 UTC · model grok-4.3

classification 💻 cs.CL cs.CY
keywords AI teaching assistantWhatsApp botscience educationWest Africafeasibility studypersonalized learninggenerative AIcurriculum-aligned
0
0 comments X p. Extension
pith:GCMG72XV Add to your LaTeX paper What is a Pith Number? 
\usepackage{pith}
\pithnumber{GCMG72XV}

Prints a linked pith:GCMG72XV badge after your title and writes the identifier into PDF metadata. Compiles on arXiv with no extra files. Learn more

The pith

An AI-powered WhatsApp bot provides accessible science learning support and assessments for West African secondary students.

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

Sub-Saharan Africa has high student-teacher ratios that limit personalized help. Adesua addresses this by running as a WhatsApp bot that lets students ask science questions and take curriculum-based tests with instant AI feedback. It draws on textbooks and decades of national exam questions to generate answers and explanations. A six-month pilot in Ghana reached 56 users and received strong ratings for usefulness from those who responded. This setup shows how existing mobile messaging can deliver low-cost, scalable tutoring where qualified teachers are scarce.

Core claim

Adesua combines generative AI with WhatsApp, curated textbooks, and 33 years of exam questions to enable students to ask questions conversationally and complete timed or untimed multiple-choice tests that receive automatic grading plus detailed feedback on right and wrong answers.

What carries the argument

The WhatsApp bot interface that merges local curriculum materials with generative AI for question answering and automated assessment.

If this is right

  • Students receive instant grading and explanations for correct and incorrect responses on science tests.
  • Support becomes available through a platform already used widely in Africa, lowering access barriers.
  • The system aligns with national curricula for JHS and SHS levels across West Africa.
  • Deployment data indicates potential for personalized learning without additional hardware or apps.

Where Pith is reading between the lines

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

  • Similar bots could extend to other subjects like math or languages if content is prepared.
  • Long-term use might allow tracking common student difficulties to guide teacher training.
  • Partnerships with ministries of education could integrate the bot into official digital learning programs.
  • Testing in other African countries would check if cultural and curriculum differences affect helpfulness.

Load-bearing premise

The small number of user ratings collected and the absence of checks on the accuracy of AI responses are enough to conclude that the bot is educationally effective and ready to scale.

What would settle it

A controlled trial that measures actual science test score improvements in students who use the bot compared to those who do not over one academic term.

Figures

Figures reproduced from arXiv: 2605.15376 by Evans Atompoya, George Boateng, Patrick Agyeman-Budu, Philemon Badu, Samuel Ansah, Samuel John, Victor Wumbor-Apin Kumbol.

Figure 1
Figure 1. Figure 1: Screenshots of Adesua misunderstood interactions. Throughout all interactions, the system maintains contextual awareness of the conversation state and responds appropriately to the user’s current position in the interaction flow. This state-based design ensures that users receive relevant prompts and that their inputs are interpreted correctly according to their current task. 3.2 Question and Answer System… view at source ↗
Figure 2
Figure 2. Figure 2: Adesua Landing Page 4 Feasibility Study and Evaluation We launched Adesua in 2025 and ran a study over 6 months. We publicized the tool on social media, sharing a catchy landing ( [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
read the original abstract

Sub-Saharan Africa faces persistently high student-teacher ratios and shortages of qualified teachers, limiting students' access to personalized learning support and formative assessment. To address this challenge, we present Adesua, a WhatsApp-based AI Teaching Assistant for science education that extends the Kwame for Science platform. Adesua leverages WhatsApp's widespread adoption in Africa to provide accessible, curriculum-aligned learning support for Junior High School (JHS) and Senior High School (SHS) students across West Africa. The system integrates curated textbooks and 33 years of national examination questions with generative AI to enable conversational question answering and automated assessment with feedback via a WhatsApp bot. Students can ask science questions, take timed or untimed multiple-choice tests by topic or exam year, and receive instant grading and detailed explanations of correct and incorrect responses. A 6-month feasibility deployment in 2025 had 56 active users in Ghana, including students and parents. Quantitative evaluation showed a high perceived usefulness, with a helpfulness score of 93.75\% for AI-generated answers, albeit with a small number of ratings (n=16). These preliminary results provide a basis for more extensive future evaluation of a WhatsApp-based AI assistant to assess its potential to offer scalable, low-cost personalized learning support and formative assessment in resource-constrained educational contexts.

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

1 major / 2 minor

Summary. The manuscript presents Adesua, a WhatsApp-based AI teaching assistant extending the Kwame for Science platform, designed to provide curriculum-aligned science learning support and automated assessment for JHS and SHS students in West Africa. It integrates curated textbooks and 33 years of national exam questions with generative AI for conversational question answering and feedback. The core contribution is a 6-month feasibility deployment in Ghana with 56 active users (students and parents), reporting a 93.75% helpfulness score for AI-generated answers based on 16 user ratings, as preliminary evidence for scalable, low-cost personalized learning in resource-constrained contexts.

Significance. If the evaluation were strengthened, this work would offer a practical demonstration of using a ubiquitous platform (WhatsApp) to address teacher shortages and high student-teacher ratios in Sub-Saharan Africa through accessible, locally relevant AI support. The integration of national exam questions and textbooks is a clear strength for alignment and relevance. As a descriptive feasibility study rather than a technical model, it provides a foundation for future larger-scale trials but currently lacks the validation needed to substantiate scalability claims.

major comments (1)
  1. The central feasibility claim—that the deployment demonstrates potential for scalable personalized learning support—depends on the reported 93.75% helpfulness score. This metric is based on only 16 self-ratings from 56 active users, with no methodology described for how helpfulness was measured (e.g., exact survey items, response scale, or collection process), no independent accuracy validation of AI answers against ground-truth textbook or exam content, and no baseline comparisons to non-AI methods or human tutors. This directly matches the stress-test concern and makes the score difficult to interpret as evidence of effectiveness or readiness for scale.
minor comments (2)
  1. Abstract: The claim of 'high perceived usefulness' should be qualified with the small rating sample size (n=16) to maintain balance and avoid potential overinterpretation by readers.
  2. Deployment description: Clarify the exact number of questions or interactions per user and any filtering applied to the 56 active users to better contextualize engagement levels.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback and for acknowledging the practical relevance of a WhatsApp-based AI assistant for science education in West Africa. We agree that the current presentation of the feasibility results would benefit from greater methodological transparency and a clearer framing of limitations to support the preliminary claims about scalability.

read point-by-point responses

  1. Referee: The central feasibility claim—that the deployment demonstrates potential for scalable personalized learning support—depends on the reported 93.75% helpfulness score. This metric is based on only 16 self-ratings from 56 active users, with no methodology described for how helpfulness was measured (e.g., exact survey items, response scale, or collection process), no independent accuracy validation of AI answers against ground-truth textbook or exam content, and no baseline comparisons to non-AI methods or human tutors. This directly matches the stress-test concern and makes the score difficult to interpret as evidence of effectiveness or readiness for scale.

    Authors: We agree that the small number of ratings and absence of detailed methodology limit the interpretability of the 93.75% figure. In the revised manuscript we will expand the Evaluation section to describe the feedback mechanism: an optional post-response prompt in the WhatsApp chat asking 'Was this answer helpful?' with a binary Yes/No scale, collected automatically from users who chose to respond. We will explicitly state that the 16 ratings represent voluntary responses from the 56 active users and note the low response rate as a limitation. We did not perform independent expert validation of AI answer accuracy against the source textbooks or exam questions in this deployment phase, as the study prioritized system integration, user engagement, and technical feasibility over controlled content auditing; the system architecture relies on curated materials to promote alignment, but we will add this as an explicit limitation and recommend such validation for follow-on work. We will likewise add a short discussion of the lack of baseline comparisons, framing the current results as preliminary evidence that motivates larger trials with comparative arms. These changes will better position the work as a feasibility study rather than a definitive effectiveness demonstration. revision: partial

Circularity Check

0 steps flagged

No circularity: descriptive feasibility study with no derivations or self-referential reductions

full rationale

The paper is a development and feasibility report on an AI WhatsApp bot for science education. It describes system integration with curated textbooks and exam questions, reports deployment metrics (56 users, n=16 ratings yielding 93.75% helpfulness), and discusses potential for scalable support. No equations, model predictions, fitted parameters, or load-bearing self-citations appear. The helpfulness score is a direct aggregate of user self-ratings, not derived from or equivalent to any internal inputs by construction. This matches the default non-circular outcome for purely descriptive studies without mathematical chains.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The feasibility claim rests on the untested assumption that generative AI outputs are sufficiently accurate for educational use and that the small user sample reflects broader adoption potential.

axioms (2)
  • domain assumption Generative AI can produce reliable curriculum-aligned explanations without systematic errors in science content for JHS/SHS level.
    Invoked when claiming the bot enables automated assessment with detailed explanations.
  • domain assumption WhatsApp is sufficiently accessible and familiar to serve as the primary interface for educational interaction in the target population.
    Stated in the abstract as the basis for accessibility.

pith-pipeline@v0.9.0 · 5800 in / 1426 out tokens · 53428 ms · 2026-05-19T15:18:26.649246+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

16 extracted references · 16 canonical work pages

  1. [1]

    https://foondamate.com/ (2026), https://foondamate.com/, ac- cessed: 2026-02-03

    Foondamate. https://foondamate.com/ (2026), https://foondamate.com/, ac- cessed: 2026-02-03

    work page 2026

  2. [2]

    Rori is an AI-powered virtual math tutor developed by Rising Academies for improving mathematics learning outcomes

    Rori: Ai-powered virtual math tutor.https://rori.ai/(2026), https://rori.ai/, accessed: 2026-02-03. Rori is an AI-powered virtual math tutor developed by Rising Academies for improving mathematics learning outcomes

    work page 2026

  3. [3]

    In: International Conference on Artificial Intelligence in Education

    Boateng, G., John, S., Boateng, S., Badu, P., Agyeman-Budu, P., Kumbol, V.: Real-world deployment and evaluation of kwame for science, an ai teaching assistant for science education in west africa. In: International Conference on Artificial Intelligence in Education. Springer (2024)

    work page 2024

  4. [4]

    Asia-Pacific Science Education9(1), 44–74 (2023)

    Chang, J., Park, J., Park, J.: Using an artificial intelligence chatbot in scientific inquiry: Focusing on a guided-inquiry activity using inquirybot. Asia-Pacific Science Education9(1), 44–74 (2023)

    work page 2023

  5. [5]

    Education and Information Technologies29(14), 18621–18642 (2024)

    Chen, C.H., Chang, C.L.: Effectiveness of ai-assisted game-based learning on science learning outcomes, intrinsic motivation, cognitive load, and learning behavior. Education and Information Technologies29(14), 18621–18642 (2024)

    work page 2024

  6. [6]

    Education and Information Technologies26(5), 6241–6265 (2021)

    Deveci Topal, A., Dilek Eren, C., Kolburan Geçer, A.: Chatbot application in a 5th grade science course. Education and Information Technologies26(5), 6241–6265 (2021)

    work page 2021

  7. [7]

    https://www.dw.com/en/why- mobile-internet-is-so-expensive-in-some-african-nations/a-55483976 (Nov 2020)

    Why mobile internet is so expensive in africa (2020). https://www.dw.com/en/why- mobile-internet-is-so-expensive-in-some-african-nations/a-55483976 (Nov 2020)

    work page 2020

  8. [8]

    Contemporary Educational Technology 17(4), ep613 (2025)

    Fayzullina, A.R., Filippova, A.A., Garnova, N.Y., Astakhov, D.V., Kalmazova, N., Zaripova, Z.F.: Artificial intelligence in science education: A systematic review of applications, impacts, and challenges. Contemporary Educational Technology 17(4), ep613 (2025)

    work page 2025

  9. [9]

    In: International conference on artificial intelligence in education

    Henkel, O., Horne-Robinson, H., Kozhakhmetova, N., Lee, A.: 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. Springer (2024)

    work page 2024

  10. [10]

    International Journal of Information and Education Technology 14(6), 876–882 (2024)

    Kurniawan, W., Riantoni, C., Lestari, N., Ropawandi, D.: A hybrid automatic scoring system: Artificial intelligence-based evaluation of physics concept compre- hension essay test. International Journal of Information and Education Technology 14(6), 876–882 (2024)

    work page 2024

  11. [11]

    Asia-Pacific Science Education9(2), 365–412 (2023)

    Lee, J., An, T., Chu, H.E., Hong, H.G., Martin, S.N.: Improving science conceptual understanding and attitudes in elementary science classes through the development and application of a rule-based ai chatbot. Asia-Pacific Science Education9(2), 365–412 (2023)

    work page 2023

  12. [12]

    Journal of Internet Technology24(2), 275–281 (2023),https://jit.ndhu.edu.tw/article/view/2867

    Lin, Y.T., Ye, J.H.: Development of an educational chatbot system for enhancing students’ biology learning performance. Journal of Internet Technology24(2), 275–281 (2023),https://jit.ndhu.edu.tw/article/view/2867

    work page 2023

  13. [13]

    In: Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP-IJCNLP)

    Reimers, N., Gurevych, I.: Sentence-bert: Sentence embeddings using siamese bert- networks. In: Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP-IJCNLP). pp. 3982–3992 (2019)

    work page 2019

  14. [14]

    International Journal of Mathematical Education in Science and Technology56(9), 1748–1777 (2025)

    Taani, O., Alabidi, S.: Chatgpt in education: Benefits and challenges of chatgpt for mathematics and science teaching practices. International Journal of Mathematical Education in Science and Technology56(9), 1748–1777 (2025)

    work page 2025

  15. [15]

    https://www.unesco.org/en/articles/persistent-teacher-gap-sub-saharan- africa-jeopardizing-education-recovery (July 2021), accessed: 2026-02-02 Adesua 11

    UNESCO: The persistent teacher gap in sub-saharan africa is jeopardizing education recovery. https://www.unesco.org/en/articles/persistent-teacher-gap-sub-saharan- africa-jeopardizing-education-recovery (July 2021), accessed: 2026-02-02 Adesua 11

    work page 2021

  16. [16]

    UNESCO, International Task Force on Teachers for Education 2030: Global report on teachers: Addressing teacher shortages and transforming the profession (2024), https://www.teachertaskforce.org/sites/default/files/ 2024-02/2024_TTF-UNESCO-Global-Report-on-Teachers_EN.pdf, see p. 50

    work page 2030