Several widely used criminal risk assessment instruments factor a defendant’s abstract beliefs, peer associations, and family relationships into their risk scores. The inclusion of those factors is empirically unsound and raises profound ethical and constitutional questions. This Article is the first instance of legal scholarship on criminal risk assessment to (a) conduct an in-depth review of risk assessment questionnaires, scoresheets, and reports, and (b) analyze the First and Fourteenth Amendment implications of attitude and associational factors. Additionally, this Article challenges existing scholarship by critiquing widely accepted but dubious empirical justifications for the inclusion of attitude and associational items. The items are only weakly correlated with recidivism, have not been shown to be causal, and have in fact been shown to decrease the predictive validity of risk assessment instruments. Quantification of attitudes and associations should cease unless and until it is done in a way that is empirically sound, more useful than narrative reports, and consistent with the First and Fourteenth Amendments.

* Attorney-Advisor, U.S. Board of Veterans’ Appeals; J.D., University of Chicago; M.P.P., University of Chicago. Thanks to Aziz Huq for guidance throughout the development and writing process; to Barbara Barreno-Paschal, David Karp, and Swethaa Ballakrishnen for their valuable comments; and to Ninet Sinaii for advice on empirical analysis. My deepest gratitude, as well, to the editors of the Stanford Law Review for their indispensable editorial work.