Ideological performance and the effects of ideological-political education are strongly associated with learners’ knowledge status and learning characteristics. This paper proposes an adaptive tracking framework that regards learners as knowledge-distillation objects and uses artificial intelligence technology as an implementation tool. The framework predicts students’ cognitive level through teacher-student cooperative extraction. On this basis, it captures prior knowledge from the teacher model and improves the predictive ability of the corresponding student model. Within this framework, the learning resource relationship model is divided into an association model, a sequence model, and a collaborative-filtering recommendation model, which are used to forecast learners’ performance on new knowledge points through multiple parameters. A hybrid differential evolutionary algorithm combined with parameter estimation is adopted to construct a personalized recommendation model for Civics learning resources based on relational patterns. The model is applied to the experimental samples’ learning process in ideological-political courses. With the support of personalized learning resources recommended by the model, the experimental group’s scores increase by 16.83 points, and the highest satisfaction rating reaches 8.1. The results show that the personalized recommendation model for ideology and politics learning resources can work together with the adaptive learning framework, taking teachers and students as dual subjects. It helps optimize and enhance students’ ideology through parameter characteristics of different knowledge points and by recommending the most compatible learning resources.
