maybe we could have a system where if the company is open source/transparent with their recommendation engine and data they are free from liability, but if they profit off the recommendation/notification content they are liable.

profit motive for recommendation/amplification is probably what damages humans at scale