The invention relates to the technical field of artificial intelligence, discloses a method for realizing individualized teaching through man-machine interaction, and provides a method for realizing individualized teaching comprising automatic decomposition of learning tasks, automatic recommendation of test questions, positioning of error steps, explanation of questions and the other teaching links through the man-machine interaction. The method has the advantages that in addition to data preparation and parameter setting in the first step, the participation of any human teachers is all not needed in other steps, thereby reducing the occupation of human resources, in particular to educational resources to the maximum extent, compared with a traditional explanation mode taking 15-20 minutes on average, by determining wrong reasons and performing step-level precise explanation, not only can content that does need to be explained be skipped to the maximum extent for specific learning conditions of a user, the questions of the user are explained emphatically, the roots of the questions are excavated deeply, so that the user can understand knowledge points more easily, but also the explanation time is shortened to 2-3 minutes, thereby greatly improving the learning efficiency of the user.