Particle.news

Michael Kremer Calls for Test-Before-Scale Policymaking in India

His Chandigarh lecture sketched how India can convert recent AI commitments into dependable results through disciplined evaluation.

Overview

  • Speaking in Chandigarh, the 2019 Nobel laureate urged officials to institutionalize an innovation cycle—identify, test, refine, then scale—so programmes expand only after rigorous evaluation.
  • He cited AI weather advisories during the 2025 monsoon that led farmers to shift nurseries, seed choices and transplanting, delivering roughly Rs 2,000 per hectare in annual gains.
  • He pointed to AI-supported driver testing now used in 56 RTOs, with data from an Ola partnership showing 20–30% fewer reported unsafe driving incidents, especially among younger drivers.
  • Drawing on a meta-analysis and Odisha trials, he said water treatment can cut under-five mortality by 21% and, if universal in rural areas, could avert over one lakh child deaths annually, provided maintenance and oversight are sustained.
  • He reported measurable learning gains from low-cost education tech—from Pratham’s SMS prompts to PAL and tablet-based teacher support—and stressed that India’s data infrastructure and 38 million AI-informed monsoon alerts show capacity to scale when usage is monitored.