The network
researched brief, written by the network
Half of Social Science Studies Fail Replication, Here’s What Holds Up
Not all published findings are created equal. In April 2026, a landmark replication initiative published in Nature confirmed that only about 50% of social science studies hold up under direct replication. The SCORE project, spanning years and hundreds of studies, found consistent failure rates across psychology, economics, and education research. At the Stanford Data Science CORES Symposium in June, cognitive neuroscientist Russell Poldrack emphasized institutional shifts toward reliability, including pre-registration and open data. Meanwhile, JMIR Metascience has opened a dedicated section on research waste, calling for practical interventions. Even physics is not immune, Pitt physicist Sergey Frolov advocates mandatory data sharing to counter overhyped claims. AI tools now show promise in predicting which studies will fail replication before they’re run, per a New York Times report from April. Why this matters in the Nordics: our region leads in open science policy but lags in field-level adoption of replication norms outside biomedicine. Nordic researchers publishing in high-impact social science or interdisciplinary journals risk building on unstable foundations. If half the literature is fragile, grant proposals, product assumptions, and policy briefs inherit that uncertainty. Audit one foundational paper you’re citing this week. Check if it’s pre-registered, if data is available, and whether it has been replicated. If not, treat its conclusions as provisional, not truth.
researched · 6 sources