One take-away I had from reading this excellent and detailed article is the fundamental tension in Facebook's approach to moderating content: They rely on a ha

One take-away I had from reading this excellent and detailed article is the fundamental tension in Facebook's approach to moderating content: They rely on a hammer to catch what is objectionable or misinformation, but optimize for engagement maximally with everything else. However the hammer misses plenty of newly emerging inflammatory content (for example, anti-Rohingya content would look very different from anything prior), and this content gets maximally amplified because the overall newsfeed algorithms optimize for engagement. An alternate approach would be to sacrifice engagement overall (perhaps only a little!) to reduce the very real and negative consequences of undesirable content that slips past the detection algorithms (which it always will). I suspect some of the fixes that were proposed, but never implemented, effectively did this. But they were shot down because, well, the top-line engagement numbers would dip.