The development of precise teaching of civics with artificial intelligence not only is the realistic need of the development of the times and technological innovation but also provides a new picture to solve the problem of the relevance of civics, so it is the inevitable requirement for the quality and efficiency of civics teaching. Curriculum Civics is an important support for the cultivation of innovation ability of college students. To effectively identify the state categories of college students’ psychology and civics and to fully consider the emotional factors between teachers and college students, this paper proposes a psychological civics teaching model based on graphical convolutional neural network. First, the dialog texts of psychology and civics teaching between teachers and students are coded in sequence context using Bi-GRU to obtain discourse text representations; then, a directed graph is constructed based on the order of the dialog between teachers and students in psychology and civics teaching, and a new text representation vector for each discourse text is obtained using graph convolutional neural network; finally, the two discourse representation vectors obtained are connected, and a similarity-based attention mechanism is used. Finally, the final discourse text representation is obtained using an attention-based mechanism to perform psychological and ideological state. The proposed method is conducive to the implementation of the teaching practice exploration of organic integration of ideals and beliefs with the cultivation of innovation ability in colleges and universities.