Yes, that makes total sense. I think detecting algorithmically would just be a pretext in my case. The real point would be to nudge the reader to ask herself and frequently: is this thing that I'm reading bullshit, and if yes, why am I wasting my time?

But if we can't get humans to agree on what is BS/Not BS... or what is fake news/not fake news, a good political view/a bad political view, moral/not moral... how in the world can we expect machines to? Even things that we can somewhat universally agree on, like what is child-appropriate content, is failing.