Thomas Lumley wrote a post about substituting LLMs for survey respondents: Let’s connect it to our discussions. Let: X = what Lumley calls a “micro-demographic” Y = vote choice Lumley first describes two problems we might see when calibrating via … Continue reading …