This chapter considers matrix-weighted consensus applied in multi-dimensional opinion dynamics and analyzes topological robustness property. Specifically, scaling matrix weights are incorporated into the existing matrix-weighted consensus and bipartite matrix-weighted consensus algorithms. Let each agent in the network have a d-dimensional state vector on d-logically interdependent topics. The introspective process of each agent is heterogeneous and being described by a positive/negative definite scaling matrix \(\textbf{S}_i\) . Two types of network interactions are considered: (i) the cooperative network containing all positive semidefinite matrix weights and (ii) a cooperative-competitive network containing both positive and negative semidefinite matrix weights. The behavior of the opinion evolution is jointly determined by the scaling matrices (private belief systems) and the matrix-weighted graphs (social interactions). The scaling matrices allow an existence of competing agents even though they share close opinion vectors. Besides generalized opinion dynamics or a scaled consensus model, a further study on bipartite matrix-weighted consensus provides insights on the performance of the matrix-weighted consensus algorithm under the sign reversals of a subset of communication signals.