An LLM Workflow That Reproduces, Improves, and Extends Published Economics Research -- by Matthew Schwartz, Isaiah Andrews, Jesse M. Shapiro
EconomicsData ScienceHigher Education
THE AI ANGLE
Auditing replication code and generating research extensionsResearchers built an open-source tool that uses language models to test 4,452 economics replication packages. The system flagged discrepancies in 3,460 papers, sped up calculations in 496, and generated new research extensions in 923. Empirical researchers and students can now run automated checks to verify and build upon published code at scale.
THE TEACHING ANGLE
The finding that an automated tool caught discrepancies in most examined papers challenges how departments teach code auditing and peer review.Read the original at nber.org Generate teaching or study materials
More in Economics
- South Korea’s AI chip giants are fighting for a bigger role in America’s AI boomRest of World · September 28, 2026
- AI in Teaching: Preserving Core Pedagogical PrinciplesFaculty Focus · September 28, 2026
- How the University of Utah built a sovereign AI factory to accelerate breakthroughs and slash cloud costsCIO.com · September 28, 2026
- Liability and Pricing of Dual-Use AI -- by Joshua S. GansNBER Working Papers · September 28, 2026
- Jensen Huang says AI distillation is 'competition.' Scott Bessent has called it 'theft'CNBC — Technology · September 28, 2026