Algorithmic Fidelity of Large Language Models in Generating Synthetic German Public Opinions: a Case Study post