Behind risk assessment, as well as most sentencing guidelines, is the assumption that arrest records are a good indicator of criminal behavior. Rather, they are partly, or mostly, indicators of how smart the defendant is (ability to escape arrest and conviction) and of whether the defendant is likely to be targeted by the police (does not look presentable, lives in a high crime area, and so on). Can any of those pushing for risk assessment point to any evidence that, among criminals, arrest records are a good indicator of likelihood to commit crime? I doubt it. Without such evidence, risk assessment should not be used.