> At least part of the ethics problem here is that it'd be plausible to conduct this research without creating any new posts.
This is a good point. Arguably though if you want people to take the next cambridge analytica or similar as something serious from the very beginning, we need an arsenal of academic studies with results that are clearly applicable and very hard to ignore or dispute. So I can see the appeal of producing a paper abstract that's specifically "X% of people shift their opinions with minor exposure to targeted psyops LLMs".
This is a good point. Arguably though if you want people to take the next cambridge analytica or similar as something serious from the very beginning, we need an arsenal of academic studies with results that are clearly applicable and very hard to ignore or dispute. So I can see the appeal of producing a paper abstract that's specifically "X% of people shift their opinions with minor exposure to targeted psyops LLMs".