Counting What Matters in Science
October 19, 2017
10/19/17 – The reward system in biomedical research can lead scientists to overlook potential biases—often unconscious—and fool themselves into believing a study’s splashy but flawed findings, a longtime science reporter recently argued in Issues, adding that such distorted studies “pervade the biomedical literature” and contribute to what’s become known as the “reproducibility crisis.” Extending this analysis, a scholar who works at the intersection of machine learning and computational linguistics says the best way around the reproducibility problem to scrap the current focus on the statistical significance of individual studies and instead pursue science in a way that “explicitly recognizes its communal and interconnected nature.”