LLMs choose friends and colleagues like people, researchers find

When large language models (LLMs) make decisions about networking and friendship, the models tend to act like people, across both synthetic simulations and real-world network contexts.