Science is supposed to be a self-correcting endeavor, but who is “the scientific expert” that corrects faulty science? We grouped traditional conceptualizations of expertise in psychology under three classes (substantialist, implicitist, and social conventionalist), and then examined how these approaches affect scientific self-correction in reference to various components of the credibility crisis such as fraud/QRPs, the inadequate number of replication studies, challenges facing big team science, and perverse incentives. Our investigation pointed out several problems with the traditional views. First, traditional views conceptualize expertise as something possessed, not performed, ignoring the epistemic responsibility of experts. Second, expertise is conceived as an exclusively individual quality, which contradicts the socially distributed nature of scientific inquiry. Third, some aspects of expertise are taken to be implicit or relative to the established research practices in a field, which leads to disputes over replicability and makes it difficult to criticize mindless scientific rituals. Lastly, a conflation of expertise with eminence in practice creates an incentive structure that undermines the goal of self-correction in science. We suggest, instead, that we conceive an expert as a reliable informant. Following the extended virtue account of expertise, we propose a non-individualist and a performance-based model, and discuss why it does not suffer from the same problems as traditional approaches, and why it is more compatible with the reform movement's goal of creating a credible psychological science through self-correction.