Social Dilemmas in Science : Detecting Misconduct and Finding Institutional Solutions

Abstract

Conditions for scientific fraud can be derived fromrational choice and game theories. Ethical norms andmoral pleas seem insufficient to prevent fraudulent behavior, as these measures can only be effective if deviations are detected and sanctioned. As a first step, effective interventions require the more frequent discovery of scientific misbehavior. To advance knowledge in this direction, we evaluate two different diagnosis tools (Benford’s lawand the caliper test), using a unique database combining 30 articles from authors in social psychology accused of committing scientific fraud, together with a control group of 30 articles published by authors who have not been the subject of suspicion.We employ Benford’s law, a common diagnosis tool for identifying fraud. The caliper test, which has commonly been used to diagnose publication bias but not outright fraud, is also used. Our results suggest that the caliper test might be a promising diagnosis tool, whereas for Benford’s law statistical preconditions are notmet. Prior applicationswithout a control group likely suffered froma too high risk of false positives. Finally, some policy recommendations are discussed.