Shaping Shared Languages: Human and Large Language Models' Inductive Biases in Emergent Communication