Paid amplification isn't the fundamental problem, but it contributes. The root of all evil on the Internet in the past 10-15 years is algorithmic timelines and algorithmic content curation whose goal function is to increase engagement.

What increases engagement? Fear, hate, conflict, and spectacle. The media has understood this for years: "if it bleeds it leads." People have a strong bias to stare at danger, probably a social adaptation to notice if a mountain lion is attacking your tribe. But when you apply it to user-generated free content, the result is the familiar "rabbit hole" and "filter bubble" phenomena and folie a deux at scale.

Back when humans directly curated content via curated lists, organic timelines that were actually just timelines, and organic non-spam-driven hyperlinking, everything was more rational and less toxic. (Even Reddit/HN style voting is better if it's truly organic and not biased by some algorithm, though it does lead to a different sort of "Borg" phenomenon sometimes.)

Even the fringe was far less toxic and more reasonable. Go find some conspiracy or other fringe topic sites from prior to 2010 on the Wayback Machine.

The problem isn't humans. The problem is that we stuck a middleman in front of human group communication and programmed that middleman to bias what messages are shown in a way that tends to amplify negativity, irrationality, and fluff as opposed to more useful and interesting but less dopamine-releasing content.