I think it has been demonstrated multiple times that toxic content increases peoples engagement i.e. time spent on the site and looking at advertising. Decreasing the time spent on the site will lower revenue. Most of what we are experiencing both on social and mainstream media is the effects of requiring an ROI for news. Before news shows had to make money (think Walter Cronkite the most trusted man in America) they actually delivered more news/facts than opinion. Today the situation is reversed. The algorithms are tuned to make this happen to maximize profit. Knowing that it is an AI doing it is irrelevant because the problem is in the model not the algorithm using the model.

> They are most likely built from existing AI frameworks, most of which are already open sourced (e.g. Tensorflow, PyTorch, Keras, etc.). Most of the mathematical equations that underpin these frameworks are published in academic journals, open to anyone who can understand them to implement, verify, and backtest

Are the algorithms actually grokable or will it just end up creating unhappiness, confusion, or more conflict. I can't imagine the algorithms take any for that will make the average person or political persona who believes the algorithm is biased happy.