This paper attempts to demonstrate the critical role of expectation horizons in economic agents building their expectations for the future. It starts with the analysis of what constraints the economics-based assumption related to information efficiency could impose in the stochastic process, and then suggests a new concept, random revision of expectation, to refer to the case when the adjustment process of expected variables employs newly generated information only. According to the inflation dynamics formula drawn under this condition, the demand pressure measured by output gap is found to cause different impacts on inflation according to different expectation horizons. The empirical analysis of this model using the data on Korea reveals that a short expectation horizon causes coefficient estimates to become small and statistically less significant.