This is really interesting. Recently I have been doing some R&D with the University of Surrey and the issue of "Discovery" ML and algorithmic behaviour has come up in discussions. The first issue raised were some of the totally bizarre and frankly awful algorithms now running on KZbin for "Discovery" To cut to the chase my wife (we live in the UK() is a HUGE fan of Netflix and we were watching Star Trek Discovery...Sorry I might be on a theme here? Brilliant series and top marks to Netflix. Then as is by magic at the end The Umbrella Academy popped back up right in our faces. As my wife had been watching one of the Umbrella Academy episodes earlier that evening kind of felt like Big Brother spitting out the screen at you. So try and build some "misbehaviour" in to your algorithmic behaviour and that way you will keep on the good side of your customers. Great service..Great shows just a mini whinge from London.....!