Psychology’s credibility crisis showed that science may dramatically fail to accomplish self-correction, at least for an extended period of time. In our opinion, one reason for this may be that the agent who will implement self-correction in science, that is, the scientific expert, is not defined with a view to scientific self-correction. According to the prevailing conception of expertise which we call the old model, an expert is an individual who has superior knowledge and skills in a particular domain. This individualist emphasis of the old model of expertise contradicts the social nature of science and the necessarily distributed nature of self-correction. The old model also undermines self-correction because it conceives expertise as something that is possessed, not performed, thus largely ignoring the epistemic responsibility of experts. We compare the old model with the extended virtue model of expertise, which defines an expert as a reliable informant in a scientific domain. A reliable informant is not necessarily an isolated individual, it can also be a research consortium or, in some respects, a whole research field. Being a reliable informant equally requires being competent and responsible, so the extended virtue model is well aligned with the goal of self-correction. We examine various issues associated with scientific self-correction such as big team science, misaligned incentives, and fraud/questionable research practices. We conclude that a revolutionary shift in how we model expertise is a must for facilitating progress in increasing the credibility of psychological science.