An Artificial Intelligence in Education Community of Interest Event
When education systems are trapped behind inaccessible and complex data systems, it becomes difficult to make informed decisions which support teachers’ and learners’ success. Our latest AI in Education Community of Interest (CoI) session, co-organised with the World Bank EdTech Team and EdTech Hub’s AI Observatory & Action Lab, brought together practitioners from research, multilateral institutions, a non-profit and the private sector to grapple with a question that sits at the heart of education governance in low- and middle-income countries: What does it actually take to use data well? And what role can AI play?
Speakers drew from hands-on experience across India, Sierra Leone, Nigeria, Bangladesh, and beyond to explore where AI adds genuine value in education data systems and where the real barriers still lie. Explore insights and discussion questions from this special session held on April 30th to learn more about this critical topic.
Featured Speakers
- Opening – Mike Trucano – Senior Advisor for Education, Technology & Innovation, World Bank
- Moderator – Krishna Kesari, Senior Digital Specialist and Data Architect, World Bank
- Carmen Strigel, Sr. Director, Center for Advanced Analytics and Visualization, RTI International; Co-Founder and CEO, Tangerine Central
- Digvijay Bhandari, Machine Learning Scientist, Wadhwani AI
- Zerin Karim, Director, AI Strategy, AI Center for Enablement, Pearson
Watch the Webinar
Key Takeaways
Here are four key takeaways from the discussion about AI and data system integration in education:
1. The challenge isn’t data scarcity, but fragmentation, access, and legibility (which AI may help address)
One of the most clarifying reframes of the session came early: the idea that LMICs are simply “data poor” is misleading and, in some ways, counterproductive. As Mike Trucano put in his opening, critical education data often exists, but lives on paper, in people’s heads, in disconnected systems, or in formats that aren’t widely accessible. Digvijay Bhandari from Wadhwani AI confirmed this and argued for the importance of consistent data standards, noting that education data is not only frequently collected and stored across disparate systems but also in inconsistent ways.
AI can play a significant role in helping make sense of these fragmented data. AI’s real affordance in these contexts – as Carmen Strigel emphasised drawing on her work across RTI International and Tangerine Central – is its ability to bring fragmented data together across systems and formats. Moreover, AI can make insights accessible, including in practical user interfaces (dashboards, chatbots, and others) for decision-makers at all levels. The barrier is not volume; it is visibility and legibility.
2. Education can learn from how other sectors have leveraged AI to maximise use of data
One of the more generative threads of the session was a question posed to Carmen about whether lessons from data integration in other sectors could be applied in education. Her answer was a clear yes.
At RTI International, AI is being used in the health sector to integrate environmental, epidemiological, and geospatial data in order to better predict zoonotic disease risks: insights that would be impossible to generate from siloed datasets alone. The parallel for education is direct. Linking attendance and learning data with community socio-economic and environmental signals, for instance, could allow systems to identify children at risk of dropping out far earlier; moving education from reactive to predictive and preventive decision-making.
Realising this potential across contexts, however, depends on how intentionally education systems design the technology and data ecosystems that underpin it. These ecosystems need strong data governance frameworks, investments in data literacy, and a commitment to keeping human judgment at the centre.
3. The technology is rarely the bottleneck: people, standards, and trust are
Zerin Karim brought a perspective grounded in both her work in the corporate education sector with Pearson and advisory work with organisations serving marginalised populations. She observed the technology is rarely the chokepoint. The tougher hurdles are capacity gaps, data standards, interoperability across teams and systems, and the absence of people who can translate between technical and local contexts.
Zerin highlighted the value of “technical translators” – individuals who speak both the technical language and the language of the communities they serve – as central to better outcomes in global AI and data projects. Digvijay echoed this, emphasising that AI is “not a magic wand” and capacity building must accompany any deployment. The group agreed that governments need clear frameworks for how data should be collected, explored, and used, as well as the allure of powerful tools distracting from the unglamorous, essential work of building that foundation. The message was blunt: don’t get stuck on what’s exciting. Sometimes the right answer is not to use AI at all.
4. Explainability should guide where AI is (and is not) applied
The session closed with a nuanced, although important, principle for practitioners and policymakers: explainability is a practical guide for responsible AI use in education data systems. Digvijay argued that where AI cannot explain what is driving a prediction or recommendation, it should not be used to make decisions affecting children, particularly when those decisions carry real consequences for learning trajectories. His oral fluency assessment example made the case concretely: e.g., AI can identify patterns in children’s reading fluency that would otherwise be invisible, and in that context its reasoning is traceable. That traceability is what makes it appropriate to use.
The group also wrestled with the persistent tension between data privacy and the incentive structures that drive data collection. Carmen raised the need to resist automation for its own sake, and to continuously monitor whether AI-enabled decisions protect learners, preserve human expertise, and avoid entrenching bias. Data enclaves — where sensitive material such as children’s audio recordings can be stored and accessed for research without personally identifiable information leaving a secure environment — were offered as one practical mechanism for navigating privacy without sacrificing insight. The broader message: data sovereignty, contextual awareness, and accountability must be treated as core design requirements from the outset, not retrofitted as compliance measures.
Further questions from the session
The following questions were posed by community members. We’re sharing to help stimulate further discussions and knowledge exchanges. Please note that some questions may have been edited for spelling or clarity.
Data quality, hallucination, and human oversight
- How do we prevent AI hallucinations when making sense of diverse and raw sets of data? To what extent do we need human control, and can we actually verify the relevance and accuracy of AI outputs?
- AI is sometimes said to help “generate” data in contexts where administrative and learning data are limited. Where can AI help fill genuine information gaps – and where should we be careful not to mistake AI-generated proxies for reliable system data?
Children’s rights and data governance
- How do you approach children’s data privacy and their legal rights? Do you agree every child has the right not to have their data harvested by AI?
- Under UK and EU GDPR, children aged 12 and under generally cannot provide their own informed consent. How do you legally obtain consent and implement the right to withdraw it?
- What does end-to-end data governance look like in practice across the full lifecycle – from collection and use to retention and deletion? Who owns decisions at each stage, and how do you ensure accountability and compliance with children’s rights throughout?
Resources
The following resources were shared by community members and participants. These have not been reviewed by the World Bank or EdTech Hub, although are useful indicators of what conversations, evidence, and methods are being explored in the sector.
Resources from the World Bank
- Data, AI, and Emerging Technologies
- Digital Public Infrastructure and Development: A World Bank Group Approach
Resources from EdTech Hub
Resources from RTI
Resources from Wadhwani AI
- Fight Against School Dropouts: Gujarat Rolls Out AI-Powered Early Warning System – AI for Social Impact – Wadhwani AI
Other resources
- UNESCO/Oxford Course on AI and Digital Transformation in Government
- UNICEF Innocenti Recommendations on Data Governance
This is part of an ongoing series hosted by the World Bank and EdTech Hub’s AI Observatory and Action Lab. The AI Observatory is made possible by support from UK International Development. Please follow along and join the conversation on LinkedIn!
