What if Recommendation Algorithms Like Facebook’s Grappled Directly with Bad Content? by Dan Frankowski | Medium

My friend Dan Frankowski has years of experience working on recommendation engines. His observations on two different models being trained for competing objectives is interesting. He also makes the great point that all algorithms have to be optimized to some outcome. Thus, any content that is algorithmically determined is optimizing for something. The only neutral methods of ordering information are in time, or random.