# worldbank-chalkboards-chatbots-genai-education-nigeria-2025-12

## Veille

World Bank: Generative AI and Education in Nigeria - RCT with Transformative Results

## Titre Article

From Chalkboards to Chatbots: Evaluating the Impact of Generative AI on Learning Outcomes in Nigeria

## Date

2025-12

## URL

https://documents1.worldbank.org/curated/en/099548105192529324/txt/IDU-c09f40d8-9ff8-42dc-b315-591157499be7.txt

## Keywords

World Bank, generative AI, education, Nigeria, RCT, Microsoft Copilot, GPT-4, English learning, AI tutoring, cost-effectiveness, learning crisis, Bloom two-sigma, Sub-Saharan Africa, EYOS, LAYS, secondary schools, Benin City

## Authors

Martín De Simone, Federico Tiberti, Maria Barron Rodriguez, Federico Manolio, Wuraola Mosuro, Eliot Jolomi Dikoru (World Bank Education Global Department)

## Ton

**Profile**: Rigorous academic public policy study, scientific and technical register, high methodological level

**Description**: World Bank research document following the strictest academic standards (RCT, multiple robustness analyses, Lee bounds, ITT/LATE). The tone is factual and measured, presenting positive results while honestly acknowledging limitations. The structure follows the classic economic working paper format. The argumentation relies on established metrics (EYOS, LAYS) enabling international comparison. Target audience: education policy decision-makers, development economics researchers, international organizations, ministries of education.

## Pense-betes

- **Study type**: Randomized controlled trial (RCT) - first rigorous study of generative AI in education in Sub-Saharan Africa
- **Intervention**: After-school tutoring program using Microsoft Copilot (GPT-4) for English learning
- **Sample**: 657 treatment / 671 control → 422/337 final, 9 public secondary schools, Benin City, Nigeria
- **Duration**: 6 weeks, 12 sessions of 90 min, 2x/week
- **Setup**: Students in pairs sharing computers, ~30 students/session
- **Main results**:
- Overall score: +0.31 standard deviation
- English: +0.23 standard deviation
- AI knowledge: +0.31 standard deviation
- Digital skills: +0.14 standard deviation
- **Equivalence**: 0.238 SD in English = **1.5 years of typical Nigerian schooling**; overall gains = 2 years
- **Heterogeneous effects**:
- **Girls**: +0.42 SD additional effect (offsets baseline gaps)
- Students with higher initial scores: larger gains
- High SES: larger effects, but low SES also significant
- **Linear dose-response**: +0.031-0.033 SD per additional day of attendance
- **Full-year projections** (21 weeks):
- At 72% attendance: 1.55 SD
- At 50% attendance: 1.2 SD
- At 100% attendance: 2.23 SD
- **Costs**:
- Per student (6 weeks): $48
- Marginal cost: $9
- Annualized: $124/student
- **Exceptional cost-effectiveness**:
- **3.2 EYOS per $100** invested (surpasses most comparable interventions)
- 0.6-1.9 LAYS per $100
- Wage returns: $7,767-$12,517 in lifetime gains (present value)
- **Benefit-cost ratio: 161:1 to 260:1**
- **Pedagogical approach**:
- 3-day teacher training
- Awareness of hallucinations and biases
- Prompts based on learning sciences (retrieval practice, elaborative interrogation)
- Supervision for on-task engagement
- National curriculum alignment
- **Scalability advantages**:
- Free software (no subscription)
- No proprietary question banks required
- Success with non-specialized staff
- Teacher as "force multiplier"
- **Global context**: 70% of 10-year-olds in low/middle-income countries cannot read a text appropriate to their level
- **Bloom's problem**: How to make the benefits of personalized tutoring accessible at scale, in a cost-effective way
- **Acknowledged limitations**:
- Randomization at student level (spillover risk)
- Urban focus (rural generalizability uncertain)
- Control group contamination in early weeks
- No comparison with human tutoring
- Infrastructure challenges (electricity, internet)

## RésuméDe400mots

The World Bank publishes the first rigorous study (RCT) evaluating the impact of generative AI on education in Sub-Saharan Africa. The intervention: a six-week after-school tutoring program using Microsoft Copilot (GPT-4) for English learning among first-year secondary school students in Benin City, Nigeria.

**Transformative results**: The study demonstrates substantial improvements despite significant infrastructure constraints. The overall score increases by 0.31 standard deviation, English by 0.23 standard deviation (equivalent to 1.5 years of typical Nigerian schooling), AI knowledge by 0.31 standard deviation. Overall gains are equivalent to two years of schooling. A linear dose-response relationship shows that each additional day of attendance generates +0.031-0.033 standard deviation of improvement.

**Notable differentiated effects**: Girls benefit from an additional 0.42 standard deviation effect, offsetting initial performance gaps. Students with higher baseline scores and those from more advantaged socio-economic backgrounds show larger gains, but disadvantaged students also achieve statistically significant improvements.

**Exceptional cost-effectiveness**: At $48 per student for six weeks ($124 annualized), the intervention generates 3.2 equivalent years of schooling (EYOS) per $100 invested, surpassing most comparable educational interventions worldwide. The benefit-cost ratio reaches 161:1 to 260:1. Projected lifetime wage returns reach $7,767-$12,517 per participant.

**Structured pedagogical approach**: The success relies on three days of teacher training, prompts designed according to learning science principles (retrieval practice, elaborative interrogation), awareness of AI hallucinations and biases, and active supervision of student engagement. The teacher acts as a "force multiplier" rather than being replaced.

**Promising scalability**: The use of free software (no subscription), the absence of any need for proprietary question banks, and success with non-specialized staff suggest strong replication potential. The study addresses Bloom's "two-sigma problem": how to make the benefits of personalized tutoring accessible at the scale of entire populations in an economically viable way.

**Critical context**: The study is set within the global learning crisis, where 70% of ten-year-olds in low- and middle-income countries cannot read a text appropriate to their level. These results position generative AI tutoring as a promising approach for resource-constrained contexts.

## GrapheDeConnaissance

- World Bank Education Global Department —publie→ Policy Research Working Paper 11125 (DOCUMENT, 0.99)
- Martín De Simone —publie→ Policy Research Working Paper 11125 (DOCUMENT, 0.99)
- Microsoft Copilot —utilise→ GPT-4 (TECHNOLOGIE, 0.99)
- Microsoft Copilot —améliore→ apprentissage de l'anglais (CONCEPT, 0.97)
- Programme de tutorat IA —mesure→ 0.31 SD d'amélioration globale (MESURE, 0.98)
- Programme de tutorat IA —mesure→ +0.42 SD d'effet supplémentaire pour les élèves filles (MESURE, 0.95)
- Programme de tutorat IA —mesure→ coût de $48 par élève sur 6 semaines (MESURE, 0.98)
- Programme de tutorat IA —mesure→ 3.2 EYOS par $100 investis (MESURE, 0.97)
- Programme de tutorat IA —résout→ Bloom two-sigma problem (CONCEPT, 0.93)
- Microsoft Copilot —est_basé_sur→ retrieval practice (METHODOLOGIE, 0.9)
- Programme de tutorat IA —observé_dans→ Benin City (LIEU, 0.99)
- crise mondiale de l'apprentissage —mesure→ 70% des enfants de 10 ans dans les PRFM ne lisent pas au niveau requis (MESURE, 0.95)
- enseignants —est_instance_de→ force multiplier (CONCEPT, 0.92)

---
Canonical: https://www.thekb.eu/en/fiches/worldbank-chalkboards-chatbots-genai-education-nigeria-2025-12/