Large language models that replace human participants can harmfully misportray and flatten identity | Xiaol.x | Podwise