Episode 79
Education at a Crossroads: AI, Equity, and Evidence
Brief description of the episode
Dan Sandhu, CEO of Education Development Trust, lays out a practical, evidence-first view of AI in education systems. He argues that AI should be deployed only when tied to clear learning outcomes, owned by ministries and school operators, and supported by evolving policies on ethics, safety, training, and leadership, with equity as the core lens. Together with host Farheen Foad, he unpacks why most “AI for schools” tools still lack serious evidence and explores how AI can support
low-bandwidth, shared-device contexts without sidelining teachers or importing hidden cultural bias.
Key Takeaways:
- Begin by checking if AI is even relevant for that ministry’s immediate challenges, since some systems face basic funding gaps that outrank AI.
- Anchor the work in real problems that ministries want solved so the deployment stays grounded in need rather than hype.
- Build programs as broader system initiatives that treat AI as only one component, rather than treating it as a standalone fix.
- Form small, structured engagements first, often through targeted system programs, before moving to wider use.
- Bring in partners with strong AI and data capability to support behind-the-scenes work.
- Set up teams on both sides that know what they are doing, so the deployment does not break due to poor expertise.
- Avoid taking on AI tasks that the team cannot deliver, as failed implementations cause more harm than progress.
- Use AI in the background to gather and process data even when classrooms have limited bandwidth or few devices.
- Provide ministries and districts with performance insights to help them act on school-quality and improvement needs.
- Give school leaders data they can use for targeted decisions without relying on heavy classroom tech.
- Support teachers by providing information that helps them deliver better lessons with minimal resources.
- Focus on simple, workable setups rather than pushing advanced classroom use when infrastructure cannot support it.
- Pace adoption slowly, so schools do not leap ahead of what their environment can realistically handle.
- Define clear outcomes and success metrics before deploying any AI tool.
- Run a trial period to confirm that the technology actually improves education, not just looks impressive.
- Gather evidence of impact before promoting or scaling the tool.
- Use AI primarily in back-office functions to support operations and decision-making.
- Combine AI with personal intervention in areas like career advisory to ensure it enables, not replaces, human judgment.
- Maintain a teacher- or advisor-centric approach, keeping humans in the loop for effective deployment.
- Work closely with ministries to understand their real needs before designing programs.
- Focus on delivering tangible educational outcomes, not just technology adoption.
- Prioritize metrics that measure learning impact over metrics that track technology usage.
- Avoid relying solely on external funding; ensure the program can operate sustainably without continuous injections of money.
- Communicate the expected outcomes clearly to all stakeholders to drive effectiveness and efficiency.
- Recognize that AI is still too new for long-term validated case studies; plan for iterative testing and adaptation over time.
- Treat pilot programs as learning phases to gather evidence before scaling widely.
Stay informed.
Subscribe to receive the latest episode in your inbox.
