UN Innovation Network 27d ago

This session explores how model behaviour varies across languages, cultures and contexts, and what this means for equitable AI adoption in low- and middle-income countries. Featuring perspectives from UNICEF, ElevenLabs, Coefficient Giving and Google DeepMind, speakers examine the limits of models trained primarily on WEIRD contexts, the role of localisation in closing rather than widening AI gaps, and how more accessible, context-sensitive tools can support inclusive adoption across diverse settings.