Paternalist policymakers face a severe knowledge problem that is analogous to the knowledge problem faced by central planners. They do not and often cannot possess the kind of local and tacit knowledge needed to craft policy interventions that reliably improve human welfare. We provide a taxonomy of types of knowledge that paternalist planners need but typically do not have: true preferences, extent of bias, self-debiasing and small-group debiasing, dynamic impacts on self-regulation, counteracting behaviors, bias interactions, and population heterogeneity. We also critique two leading efforts to surmount knowledge problems of behavioral paternalism: the augmented revelatory frame approach and unified behavioral revealed preference.

The policies recommended by behavioral paternalists rest on certain posited empirical facts or regularities about human behavior. Some of these supposed facts have not been established with much confidence. Specifically, psychological findings are highly context-specific, and thus lack the generality required for policymaking; generalizing quantitative results from the laboratory to the real world is unreliable; most existing research does not account adequately for incentives and learning; most existing research does not consider small-group debiasing; and most existing research does not adequately assess self-regulation and self-debiasing. Establishing reliable answers to these questions is a prerequisite for crafting and calibrating paternalistic policies with a reasonable expectation of improving welfare.