Abstract: 【Purpose/significance】Exploring the formation of attitudinal ambivalence among users on social media, gaining a deeper understanding of the user dynamic behavior changes in social media, and exploring social media user category tags and behavioral characteristics have become urgent issues for the continuous operation of social media.【Method/process】This research constructs the generation mechanism and framework model of social media users' attitudinal ambivalence, collects 586 sample data through question⁃ naires, and conducts cluster analysis and calculation and determination of the degree of attitudinal ambivalence on individual cogni⁃ tion, social environment, positive experience, negative experience, positive emotion, negative emotion and other factors. Divide users into four categories based on their level of attitudinal ambivalence: the most high, high, medium, and low, and compare the differences in labels and behavioral performance of users with different levels of attitudinal ambivalence.【Result/conclusion】Through analysis, it was found that the influencing factors of attitudinal ambivalence can lead to varying degrees of attitudinal ambivalence and behavioral manifestations on social media, and there are significant differences in the degree of influence, providing theoretical support for further optimizing social media platforms.【Innovation/limitation】Introducing attitudinal ambivalence into the study of social media user be⁃ havior and empirically discover the dynamic changes in user behavior with different attitudinal ambivalence. However, further explora⁃ tion is needed for dynamic monitoring of user behavior data.