Launch event on Building Evidence in Education’s (BE²’s) new working paper on responsible AI in education evidence

, 

On Wednesday, 4 February 2026, Building Evidence in Education (BE²) hosted a public webinar to launch our new working paper, Artificial Intelligence in Education Evidence Synthesis and Use in Low- and Middle-Income Countries. The session brought together researchers, evidence intermediaries, funders, and practitioners to focus on a practical question: How can AI responsibly speed up evidence synthesis and evidence uptake, without weakening trust, quality, equity, or local evidence ecosystems?

If you missed it, we encourage you to watch the recording and download the paper and executive summary.

The working paper provides practical guidance on when AI can add value and when it should not. It outlines the quality assurance and human oversight needed to use AI responsibly, with a focus on ensuring AI strengthens LMIC evidence ecosystems rather than reinforcing bias, inequities, and dependency.

Highlights from the Discussion

  • Deborah Greebon (BE² Secretariat) introduced the paper’s core ideas and recommendations, including where AI can responsibly support evidence synthesis and evidence use, and where risks are highest.
  • Dr. Ezequiel Molina (World Bank) offered a key reframing. AI is not only making evidence work cheaper. It is changing the nature of the product we should be building. He argued for investing in a different kind of evidence syntheses that are designed for how evidence is actually used today, and he challenged the field to build stronger benchmarking and protocols as AI tools evolve.
  • Jonathan Kay (Education Endowment Foundation) spoke directly from EEF’s evidence synthesis practice, including experience informing the Teaching and Learning Toolkit. He emphasised iterative development, testing, and verification. He also argued for using AI to support a systematic process, rather than trying to replace it, and noted that high-performing workflows are often bespoke rather than off-the-shelf. He highlighted collaboration as one of the most important near-term actions, given how many organisations are working on overlapping technical problems.
  • Dr. Sreehari Ravindranath (Dream a Dream, India) grounded the discussion in work underway through Dream a Dream’s GenAI-supported Research–Policy Lab and its Collaborative Data Ecosystem. His reflections emphasised that evidence uptake is deeply relational. AI may lower access costs, but it cannot replace professional judgment. He also highlighted AI’s potential value as a shared object that supports collective sense-making across stakeholders.
  • Dr. Madiha Khan (Educate Ventures Research) moderated the session and guided the conversation across perspectives, drawing out practical implications for funders, evidence intermediaries, and decision-makers working in LMIC contexts.

The strong turnout and level of engagement underscored how much interest there is right now in this topic.

Want to collaborate?

BE² is keen to connect with funders, evidence intermediaries, researchers, and government partners who are exploring responsible AI for evidence synthesis or evidence uptake. If you are interested or willing to share what you are learning, please get in touch at be2_admin@building-evidence-in-education.org.