What are the odds that the Twitter accounts that the researchers base their findings on were also bots? It’s an interesting thought to entertain with intertwined mutual feedback loops of assisted automated systems all assuming each other to be human.
Yes, I didn't see any indication from the article that they verified that the users in question were human, or that they were not paid to tweet misinformation. The effect was pretty small, it seems like even a small number of dedicated bad actors could account for it.
