Why can permutation importance underrate correlated features?

Question: A model relies on either of two strongly correlated variables interchangeably. Why might ordinary one-feature-at-a-time permutation importance assign both variables low importance? A) Permutation importance is calculated only from tree split counts B) Permutation preserves every conditional dependency involving the shuffled variable C) When one variable is shuffled, its correlated substitute can preserve much of the model's predictive performance D) Correlated variables necessarily receive identical permutation-importance values Correct: C Explanation: Permuting one correlated feature may not substantially damage prediction because another feature carries similar information. Consequently, both variables can appear individually unimportant even when the correlated group is highly predictive. Topic: advanced ML / interpretability / permutation importance