In line with this year's theme Evaluation, Evidence, and Trust in the Age of AI, this panel brings together members of the EvalforEarth Community of Practice to examine how evaluation can remain credible, inclusive, and context-responsive in an AI-augmented world. Drawing on experience from Nepal, and Georgia, the session offers a cross-contextual reflection on the intersection of policy, practice, and methodology in the use of artificial intelligence for evaluation across food security, agriculture, and rural development.
The panel presents complementary perspectives. Ramesh (Nepal) explores the implications of Nepal's National AI Policy 2025, the country's first for development evaluation systems, governance, and public trust, drawing lessons for countries navigating responsible AI adoption under real infrastructure and data constraints. Dea (Georgia) uses Outcome Harvesting as a concrete methodological test case to examine where AI can responsibly support evaluation practice and where human judgement remains non-delegable, bridging field experience across multiple country contexts with reflections on evaluator training at the University of Tbilisi.
These voices make the case that maintaining trust in AI-supported evaluation depends not on technological innovation alone, but on human-centred approaches, ethical safeguards, and the capacity of evaluators to critically and deliberately navigate AI's role in diverse and rapidly evolving contexts.