Persuasiveness of generative AI models
How well can Generative AI models like ChatGPT persuade people and change their minds?
Karinshak2023working studies AI model’s ability to generate pro-vaccine messages.
Mor Naaman worked on a couple of papers on this line.
Costello2024durably argues that it is possible to convert some true believers of Conspiracy theories through a conversation with LLMs.
What just happened, what is happening next by Ethan Mollick
Hackenburg2024evidence reports a log-Scaling law for persuasive power. Hackenburg2024evaluating found that GPT is quite persuasive but microtargetting doesn’t seem to work better than non-targeted messages.