None
EN
Using AI to Engage People about Conspiracy Beliefs
['Justin Hendrix']
The Tech Policy Press Podcast
If you say that people above 50% believe it, people below 50% don't believe it, of our initially 100% people believing it, 25% of them didn't believe it after the conversation. And I think that to me, the biggest problem with LLMs and content moderation, or let's say one big problem with LLMs and content moderation, is unlike conspiracy theories where, these classic conspiracy theories, a huge amount has been written about them on the internet and the conspiracy theory, sorry, the LLMs have a sort of deep fund of knowledge about JFK assassination, for example, if it's some piece of misinformation about current events, like the LLM is not going to know about it.