Manuel Tonneau

LLM demographic prompting shows bias, inconsistency

New research indicates that using sociodemographic prompting with large language models (LLMs) to align their judgments with specific demographic groups can lead to unintended consequences. One study found that models w…