Do LLMs Reproduce the Age Gap in Loneliness? Evaluating Survey Simulation against European Survey Data
We test whether LLM outputs reproduce within-country age gaps in loneliness using the 2022 EU Loneliness Survey, comparing adults aged 18-34 and 65 or older in France, Italy, Poland, and Sweden.
ProofPaper ↗
Key points
- Large language models (LLMs) are often used to simulate survey respondents, but matching survey averages does not establish accurate demographic differences.
- We evaluate simulation with and without age information and direct distribution estimation against a probability-based sample, with a non-probability sensitivity reference.
- Both human samples show higher mean loneliness among younger adults.
- For DeepSeek, simulation has lower gap error but direct estimation has lower group-mean error; these comparisons depend on low-validity older-adult simulation cells.
Sources (1)
- [1]Do LLMs Reproduce the Age Gap in Loneliness? Evaluating Survey Simulation against European Survey DataarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 18, 10:31 AM
We test whether LLM outputs reproduce within-country age gaps in loneliness using the 2022 EU Loneliness Survey, comparing adults aged 18-34 and 65 or older in France, Italy, Poland, and Sweden.
Large language models (LLMs) are often used to simulate survey respondents, but matching survey averages does not establish accurate demographic differences.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 17, 2026A Cross-Domain Evaluation of Prompting Techniques for Large Language Models
- Oct 17, 2026Medical Fine-Tuning Reshapes, Not Reduces, Patient-Facing AI Risk: A Rubric-Based, LLM-Judge Evaluation Framework
- Oct 16, 2026Multi-author style detection with stylistic embeddings
- Oct 16, 2026Characterising and Analysing Retrieval-Augmented Generation Techniques
- Oct 16, 2026NL2SQL with Iterative Self-Refinement
- Oct 16, 2026Language Models Learn to Improve Language Models