The research examines the accuracy and fairness of commercial risk assessment software, COMPAS, which is used to predict recidivism. The study compares the predictions made by COMPAS with those of untrained individuals who participated in an online survey. Despite the complexity and extensive features of COMPAS, the study finds that the predictions made by the untrained participants are nearly as accurate and fair as those made by COMPAS. Specifically, a simple linear classifier using only two features (age and total number of previous convictions) achieves similar prediction accuracy to COMPAS. The study also shows that more sophisticated classifiers, such as a nonlinear support vector machine, do not significantly improve prediction accuracy or fairness. These findings cast doubt on the effectiveness of algorithmic recidivism prediction and suggest that the current methods may not be necessary for making informed decisions in criminal justice.The research examines the accuracy and fairness of commercial risk assessment software, COMPAS, which is used to predict recidivism. The study compares the predictions made by COMPAS with those of untrained individuals who participated in an online survey. Despite the complexity and extensive features of COMPAS, the study finds that the predictions made by the untrained participants are nearly as accurate and fair as those made by COMPAS. Specifically, a simple linear classifier using only two features (age and total number of previous convictions) achieves similar prediction accuracy to COMPAS. The study also shows that more sophisticated classifiers, such as a nonlinear support vector machine, do not significantly improve prediction accuracy or fairness. These findings cast doubt on the effectiveness of algorithmic recidivism prediction and suggest that the current methods may not be necessary for making informed decisions in criminal justice.