Potential Human Rights Violations in Risk Assessment of Artificial Intelligence Systems: Two Cases of Recidivism Prediction Tools