Computationally expensive — exact Shapley values require evaluating all feature subsets, which grows exponentially with the number of features - Approximations (KernelSHAP, TreeSHAP) are faster but introduce their own assumptions - Feature interactions can be missed — Shapley values attribute import