Social media is becoming a crucial communication tool for information generation, dissemination, and consumption. Can evidence drawn from social media enhance public services and inform public policy development? In this talk, we present studies using billions of geotagged social media posts and state-of-the-art natural language processing techniques to provide valuable insights for researchers and policymakers about global preferences. We present three case studies on air pollution, climate change, and COVID-19. These applications show how social media data, coupled with machine learning techniques and quasi-experimental design, can inform public decision-making.