The decision-making component in a mixture-of-experts model that determines which experts should process each input token.