Fig. 1 gives an example of a conversation where the user goes from 100% belief down to 40% after getting their questions explained by the AI.
Looking at the conclusions, the impact is not so big for all the interactions.
Anyway, this is a great tool. Sure, when people are doomscrolling 24/7 they’re not fact-checking. So, the intervention might not be there. Yet, I choose to remain optimistic. More recent generations might get easier access and be better than our current trend :)
In before conspiracy freaks start inventing conspiracy theories about everyone who tries to talk them down being a malicious AI.