In this talk two important applications of meta-induction are developed. The first is the application to probability aggregation in Bayesian prediction games. In these games the predictive targets are probability distributions scored by the truth-values of the events. Meta-inductive probability aggregation achieves a predictively optimal distribution among all available probabilistic methods. The second application relates to social epistemology. Here here a form of social learning called local meta-induction is investigated. In local meta-induction it is assumed that individuals can access only the success records of the individuals in their immediate epistemic neighborhood. It is shown that local meta-inductive learning can spread reliable information over the en-tire population, and has clear advantages compared to success-independent social learning methods such as peer-imitation and authority-imitation.