Between Hype and Harm: Learnings from Global South experiences of using AI in Evaluation

Artificial intelligence (AI) is increasingly being used across the evaluation cycle from transcription and translation to qualitative analysis and reporting, promising efficiency while raising ethical, methodological, and trust-related concerns. These tensions are particularly pronounced in Global South contexts, where evaluations often take place in low-connectivity, fragile, or politically sensitive environments.
This panel brings together evaluators from South Asia and other Global South contexts to share practical, experience-based learnings on using AI responsibly in evaluation practice. Rather than advocating uncritical adoption or outright rejection, the discussion focuses on real-world dilemmas: fear of scrutiny for using AI, lack of institutional guidance, ethical grey areas, and risks to data quality, confidentiality, and community trust.
Panelists will present concrete examples of AI use across data management, qualitative analysis, sensemaking, and reporting, highlighting where AI-supported rigor and where human judgment were essential to avoid bias or misinterpretation, particularly in humanitarian, governance, gender, and refugee response evaluations.
Grounded in core evaluation principles such as validity, transparency, informed consent, and contextual sensitivity, the session also examines how power, gender, and inclusion shape AI impacts. Through audience interaction, participants will co-develop a practical AI in an Evaluation Decision Tree to help distinguish tasks suitable for AI support from those requiring human judgment.