Moderate evidence
More
Review of 300 sources finds structured AI tutoring helped students in Nigeria and Ghana
Study title: Towards a Pedagogical Framework for Integrating Generative AI Into Competency-Based Education in Secondary Schools in the Global South
- Evidence type
- Systematic review or meta-analysis
- Sample
- n = 300
- Sources
- 1 sourcePeer-reviewed
- Published
- 2 min read
- Truvace record
- TRV-2026-3071cite or check it
Study title: Towards a Pedagogical Framework for Integrating Generative AI Into Competency-Based Education in Secondary Schools in the Global South
A systematic review of 300 sources, drawn from 12,125 Scopus and Web of Science records published between 2015 and 2025, examined how generative AI could support competency-based education in secondary schools across the Global South. It found that randomized trials in Nigeria and Ghana reported meaningful learning gains when AI tutoring was structured and supervised.
Read the study abstract
Secondary education systems across the Global South have adopted competency-based curricula that expect learners to construct knowledge, solve authentic problems, collaborate and demonstrate skills, yet large classes, teacher shortages and limited resources constrain the learner-centered pedagogy that these reforms require. Generative artificial intelligence (GenAI) has been proposed as a means of closing this gap, but its pedagogical role in competency-based education has not been systematically theorized. Guid…
The gain
A systematic review of 300 sources found that randomized trials in Nigeria and Ghana reported meaningful learning gains when generative AI tutoring was structured and supervised.
In brief
The review also found that students given unguided access showed overreliance and weaker independent performance. Competency-based reforms were already held back by teacher preparation, assessment practices and limited resources, which is the gap AI has been proposed to fill.
Connectivity, language, cost, teacher capacity and policy gaps limited equitable use. Building on the findings, the authors propose a six-phase framework that fades AI support over time, but it is a proposal and has not been tested in classrooms.
Main points
- The review covered 300 sources selected from 12,125 Scopus and Web of Science records published between 2015 and 2025, reported under PRISMA 2020.
- Randomized trials in Nigeria and Ghana reported meaningful learning gains when generative AI tutoring was structured and supervised.
- Unguided access to generative AI was associated with overreliance and weaker independent performance.
- Connectivity, language, cost, teacher capacity and policy gaps constrained equitable use in resource-limited schools.
- The authors propose the SGAI-CBE framework, a six-phase cycle with teacher-mediated scaffolding and gradual fading of AI support.
The full story
Secondary education systems across the Global South have adopted competency-based curricula that expect learners to construct knowledge, solve authentic problems and collaborate. Large classes, teacher shortages and limited resources constrain the learner-centered teaching these reforms need, and the review found reforms were held back by teacher preparation, assessment practices and resources.
Guided by Vygotsky's social constructivist theory, the authors found generative AI offered support for scaffolding, dialogue, feedback, authentic task design and teacher support. The strongest outcome evidence came from randomized trials in Nigeria and Ghana, where gains appeared under structured and supervised tutoring.
From these findings the authors propose the Scaffolded Generative AI for Competency-Based Education framework. It is anchored in the zone of proximal development, teacher-mediated scaffolding, collaborative knowledge construction, authentic assessment and the gradual fading of AI support, within guardrails suited to resource-constrained schools.
Sources
1 on file- Educational Research and Developmentdoi:10.65343/erd.v2i4.171
Talk it over
The randomized trials in Nigeria and Ghana found gains under structured and supervised AI tutoring, while unguided access went with overreliance and weaker independent performance. How would you weigh these two findings against each other?
Review of 300 sources finds structured AI tutoring helped students in Nigeria and Ghanahttps://truvace.com/item/towards-a-pedagogical-framework-for-integrating-generative-ai-into-competency-ba
The randomized trials in Nigeria and Ghana found gains under structured and supervised AI tutoring, while unguided access went with overreliance and weaker independent performance. How would you weigh these two findings against each other?
Answer in the threadThe review drew 300 sources from 12,125 Scopus and Web of Science records, yet the strongest outcome evidence came from trials in Nigeria and Ghana. What can those trials tell us about schools elsewhere in the Global South?
Answer in the threadConnectivity, language, cost, teacher capacity and policy gaps limited equitable use. Which students in resource-limited schools might be missing from the trial gains, and why?
Answer in the thread
If you have more time
The authors propose a six-phase framework that fades AI support over time but has not been tested in classrooms. What would a fair classroom test of it look like, and who should run it?
Answer in the threadThink of a school you know of. Which of the gaps the review names (connectivity, language, cost, teacher capacity, policy) would matter most there, and what would you check first?
Answer in the thread
ace
What readers say