Molecular dynamics simulation is applied to investigate the mechanism and variation of self-diffusion in calcium aluminosilicate slags. The self-diffusion coefficients are calculated for eleven slag compositions with varying Al₂O₃/SiO₂ ratios at a fixed CaO content. In practice, the results of the study are relevant to the significant changes in transport phenomenon caused by the changes in chemical composition during continuous casting of steels containing high amounts of dissolved aluminum. The cooperative movement between O atoms and network formers is discussed since [AlO₄] and [SiO₄] tetrahedra are the elementary structural units in the CaO-Al₂O₃-SiO₂ (CAS) slag system. The diffusivities for four atomic types are affected by the degree of polymerization (DOP) of slag network characterized by the proportions of non-bridging oxygen (NBO) and Qⁿ species in the system. On the other hand, a sudden increase in 5-coordinated Al as network modifiers in high alumina regions slightly increases the self-diffusion coefficient for Al. As another structural defect, oxygen tricluster plays an important role in the behavior of self-diffusion for O atoms, while the diffusivity for Ca is deeply influenced by its bonding and coordinating conditions.