Most major large language models (LLMs) can quickly tell when they are being given a personality test and will tweak their responses to provide more socially desirable results-;a finding with implications for any study using LLMs as a stand-in for humans.
