PSYCH Dataset - Psychometric Assessment of Large Language Model Characters: An exploration for the German language

This dataset contains large‑scale psychometric response data generated by 32 contemporary large language models (LLMs) evaluated using the German Big Five Inventory‑2 (BFI‑2). Each model completed all 60 BFI‑2 items 60 times, under controlled conditions including male and female persona impersonation and responses given with and without short justifications, resulting in over 330,000 valid responses. The dataset enables systematic comparison of LLM‑generated personality‑like response patterns with validated human reference data from the German BFI‑2 norm sample. It supports analyses of behavioral consistency, justification effects, and gender‑related bias in non‑English language contexts. Data are provided in structured, machine‑readable formats to facilitate reproducibility and reuse in psychometric and AI evaluation research.