It’s somewhat more nuanced than that; a classic sampling problem. When Instagram knows nothing about user X, it trends to shocking content that user X may not

It’s somewhat more nuanced than that; a classic sampling problem. When Instagram knows nothing about user X, it trends to shocking content that user X may not want. Which is bad. But the reason it does so is that the shocking content has higher engagement, in average, among users A - W. It is not right to blame user X, and it is very likely that users D, H, and M are dismayed at the shock content but can’t help slowing down when they scroll past it. So we can blame Meta to some degree, and users A - W (except D, H, and M) to some degree, and users D, H, and M to some degree. But if the total ad revenue from users A-W was lower when shock content was featured, user X would not be presented with it. Of course, Meta could just be decent. But so could the other users. The fact that neither are true is an emergent property of sociopathic corporations and an unfortunate side of human nature.