Replacing human research subjects with AI chatbots.
Open link in next tab
Just a moment...
https://www.science.org/content/article/can-ai-chatbots-replace-human-subjects-behavioral-experiments
As far as I understand this, they seem to think that AI models trained on a set of affluent westerners with unknown biases can be told to "act like [demographic] and answer these questions."
It sounds completely bonkers not only from a moral perspective, but scientifically and statistically this is basically just making up data and hoping everyone is impressed by how complicated the data faking is to care.