●researcherThis is a direct validity challenge to using LLMs as synthetic social media user simulators — the sensitivity finding should inform experimental design in any study relying on LLM-simulated opinions.
●policyWorth watching because fragile stance simulation undermines the reliability of LLM-based tools used for content moderation research or public opinion modeling.