[GraffitiTim]

It would be interesting to see ML recommendations of the form "There is a 75% chance you'll LOVE this movie, 25% chance you'll hate it" rather than only "98% chance you'll like this movie". Right now those 75% movies are actively hidden, as 75% is considered a bad prediction.

In other words, rather than optimizing for zero risk of disliking something, allow the user to optimize for the possibility of truly loving something, or being surprised. Especially on Netflix, where the risk is you stop part way through and switch to something else!