And you don't have to be a conspiracy theorist who thinks that the heads of Twitter cynically tuned the algorithm to reward anger. If it's human nature to love a good fight, then unguided ML would generate a black-box model that promoted angry posts.
I never really got those conspiracy theories. The big secret is they're tuning it to make money. Money in the ad-tech age of the internet means eyes on content and engagement. Anger and clickbait and such are all incredibly easy to tune for. Plus it's not just the algorithms that are tuned for it, people, including the "creators" on any platform, change their behavior to do "better" on the platform.
It's just a feedback loop optimized for "engagement" which is the metric these companies are using to make more money from the companies buying ad-space. Occam's razor and all that. It's just about money.