AI chatbots can fake survey responses, but they miss the human weirdness

Researchers used ChatGPT to simulate how people respond to surveys about payment apps. It turns out the AI can mostly match real human answers, especially on privacy concerns. But it cannot replicate the full range of human variability or the subtle ways people express their actual concerns.

Companies and governments use surveys to understand public opinion on new technologies. If AI can generate plausible survey responses, it could make it harder to tell what real people actually think. This means that relying on AI for market research or policy feedback could lead to decisions based on artificial consensus, not genuine public sentiment.