Prompt Perturbations Reveal Human-Like Biases in Large Language Model Survey Responses

Open in new window