Half of social-science studies fail replication test in years-long project
6 hours ago
- #research reproducibility
- #open science
- #social science
- A seven-year SCORE project found only half of 3,900 social-science papers could be replicated, highlighting widespread issues in reproducibility and robustness.
- Key challenges included insufficient data or details in papers, leading to difficulties in reproducing analyses and experiments, though automation tools may help improve future research.
- Independent research suggests newer studies (2022–23) show better reproducibility rates (85%) due to improved data and code disclosure norms, indicating progress in research transparency.