How does the brain control its sensory plasticity using performance feedback? We examined this question using various types of fake feedback in perceptual learning paradigm. We demonstrated that fake feedback indicating a larger performance improvement facilitated learning compared with genuine feedback. Variance of the fake feedback modulated learning as well, suggesting that feedback uncertainty can be internally evaluated. These results were explained by a computational model which controlled the learning rate of the visual system based on Bayesian estimation of performance gradient incorporating an optimistic bias. Our findings suggest that sensory plasticity might be controlled by high-level cognitive processes.