Artificial intelligence risks amplifying inequities in education unless countries address fundamental barriers to teaching and learning, delegates heard at a roundtable hosted at the University of the Witwatersrand in Johannesburg on Aug. 5.
Delivery, not curriculum, is the chokepoint
Speakers argued that the central challenge is not what is taught but whether schools can actually deliver instruction — a point with direct relevance for policymakers and school systems wrestling with digital tools.
Kume Luvhani, co‑founder and executive director of Vaxowave, told the session that artificial intelligence will not automatically equalize education. Instead, he said, it will strengthen whatever system it is introduced into.
“AI will not equalize African education by default. It will amplify the system it enters.”
He emphasised that simple connectivity is not the same as meaningful access. That fuller notion of access, he said, requires several layers — from reliable infrastructure and affordability to trained teachers and ongoing support — and without them AI may only make disparities more sophisticated.
Practical barriers overshadow policy ambition
Speakers highlighted persistent gaps that undermine implementation even where strong policy exists: insufficient laboratory facilities, limited bandwidth and a shortage of educators trained to use advanced tools. In short, the issue is often one of capacity and capital rather than curriculum design.
The roundtable was organised by AI ethics group AITHOS together with Wits Machine Intelligence and Neural Discovery Institute and brought together educators, researchers and industry practitioners to discuss how AI should be governed and scaled.
Participants underscored governance as a crucial consideration — who makes decisions about AI in schools, and whose perspectives are included. The event placed particular emphasis on gender and generational representation in those conversations.
“The people who will live longest with today’s decisions are largely absent from them,”
The quote, attributed to Aashna Jain, founder and chief executive of AITHOS, was used to argue for including young people in the design and oversight of AI systems given that they will experience long‑term consequences.
What meaningful access entails
Speakers unpacked the components that, combined, create meaningful access to AI in education. They listed them as:
- Infrastructure — reliable electricity and connectivity;
- Affordability — cost models that schools and families can sustain;
- Capability — teacher training and curriculum integration;
- Safety — protections for learners and data privacy;
- Ongoing support — maintenance, updates and pedagogical coaching.
Those elements were presented as prerequisites for AI to act as an enabler rather than a force that entrenches advantage.
| Component | Role |
|---|---|
| Infrastructure | Enables consistent use of digital tools in classrooms |
| Capability | Ensures teachers can effectively integrate AI into pedagogy |
Implications for policy and practice
The discussion at Wits serves as a reminder for education leaders in Canada and elsewhere: introducing advanced technology without parallel investments in physical resources and human capacity risks widening disparities. Policymakers must therefore balance ambition with a clear plan to support schools and educators on the ground.
Speakers urged that governance arrangements be widened to include younger voices and women, groups they said are often underrepresented in decisions that will shape learning environments for years to come. That call reflects a broader concern about who sets priorities for educational technology and whose needs are centred.
As jurisdictions consider pilots or scale‑ups of AI tools, the roundtable’s message is straightforward: scale opportunity, not inequality. That requires targeted spending on infrastructure, investments in teacher professional development, and governance frameworks that include those most affected.
For students, families and educators, the challenge is practical and immediate. Without basic supports, advanced software and algorithms will be tools in search of a context, not instruments of improved learning.
