Prof Nick co-leads groundbreaking study on AI and global well-being

Key findings show that while LLMs can identify broad correlates of well-being, such as income and health. However, their accuracy declines in underrepresented regions, highlighting biases linked to global digital and economic inequalities. The study also finds that these models often rely on surface-level language cues rather than deeper contextual understanding, which can lead to misestimations in unfamiliar contexts.

Incorporating data from underrepresented regions does improve predictions, though gaps remain. LLMs cannot yet replace human-collected data for predicting well-being, and continued collection of high-quality human data remains essential. We look forward to sharing more highlights on NTU’s growing impact in globally relevant, cutting-edge AI and social science research.

Click here to read the study.