One Persona, Many Cues, Different Results: How Sociodemographic Cues Impact LLM Personalization
This addresses bias and fairness issues in LLM personalization for researchers and developers, but it is incremental as it builds on prior persona-based studies.
The study examined how different sociodemographic cues (e.g., user names) affect LLM personalization across tasks and models, finding that while cues are correlated, they lead to significant response variance, cautioning against relying on single cues.
Personalization of LLMs by sociodemographic subgroup often improves user experience, but can also introduce or amplify biases and unfair outcomes across groups. Prior work has employed so-called personas, sociodemographic user attributes conveyed to a model, to study bias in LLMs by relying on a single cue to prompt a persona, such as user names or explicit attribute mentions. This disregards LLM sensitivity to prompt variations (robustness) and the rarity of some cues in real interactions (external validity). We compare six commonly used persona cues across seven open and proprietary LLMs on four writing and advice tasks. While cues are overall highly correlated, they produce substantial variance in responses across personas. We therefore caution against claims from a single persona cue and recommend future personalization research to evaluate multiple externally valid cues.