Study with CEP survey data tests surveys answered by artificial intelligence: They work with less accuracy in Chile than in the U.S.
The research, published in EPJ Data Science, evaluated the silicon sampling technique, which replaces real survey respondents with AI-simulated responses. In Chile, the models showed lower accuracy for women, older adults, religious people, those with lower educational levels, and lower-income groups. Large language models (LLMs), such as those behind ChatGPT, Gemini, or Claude, began a…