People often use what they know as a basis to estimate what others know. This egocentrism can bias their estimates of others' knowledge. In 2 experiments, we examined whether people can diminish egocentrism when predicting for others. Participants answered general knowledge questions and then estimated how many of their peers would know the answers. Egocentrism was revealed in the relationship between participants' own accuracy and their estimates of peer accuracy for questions that were new to the experiment. However, when participants encountered the answer to a question asked earlier in the experiment, they showed reduced egocentrism for these old relative to new questions (Experiment 1). Participants were aware that recent experience with answers spoiled their knowledge as a basis for estimating what others know. Consequently, they relied on more objective bases for prediction, which enhanced their ability to discriminate between questions that are easy versus difficult for others (i.e., relative accuracy). In Experiment 2, the relative accuracy of estimates of others' knowledge was also enhanced when experience-based cues were blocked by presenting the answer with the question. Results are discussed in terms of a dual process theory of the bases (e.g., experience vs. theory) people use for predictions for others. Further, we discuss the effects of egocentrism in educational contexts, such as a professor estimating what students know. In sum, our findings show that people can shift away from their own knowledge to diminish egocentrism and to more accurately estimate what others know. (Contains 4 tables, 5 figures, and 4 footnotes.)