German General Social Survey Personas: A Survey-Derived Persona Prompt Collection for Population-Aligned LLM Studies

Researchers have created GGSS Personas, a comprehensive collection of survey-derived persona prompts based on the German General Social Survey that helps Large Language Models simulate human perspectives more accurately. The collection enables LLMs to generate responses aligned with the German population and outperforms existing classifiers, particularly when training data is limited.

• →GGSS Personas is a new persona prompt collection built from German General Social Survey data to improve LLM human simulation accuracy.
• →The persona-guided LLMs outperformed state-of-the-art classifiers, especially in data-scarce scenarios.
• →The collection is designed to be easily integrated into various LLM prompts and tasks across different models.
• →Researchers evaluated how representativity and attribute selection in persona prompts affect population alignment.
• →The resource enables more systematic exploration of population-aligned persona prompting in NLP and social science research.