The mechanism that decides which specialized sub-networks (experts) should process each input in a mixture-of-experts model.