Over the last decade, in a major switch in position, conservatives have embraced the cause of reducing prison populations in the states and, increasingly, at the national level. The long-term crime decline and the increasing antistatism of the Republican Party contributed to this change, but it also has an important cognitive component: Policy makers have become more open to evidence of the damaging effects of mass incarceration. In contrast to previous studies, our case shows that such policy “feedback” only functions politically when a signal about a policy consequence is assigned valence and intensity by policy makers, whose calculations are heavily structured by the demands of party coalitions. On issues in which no core coalition member has a major stake, feedback can be tipped from reinforcing to undermining and vice versa, but this process depends on the efforts of entrepreneurs to change the way information is processed. In a highly polarized environment, opening policy makers to previously ignored evidence requires the cultivation of a reform cadre composed of ideological standard-bearers who can vouch for the orthodoxy of the new position.