> I think all data can tell us is what people _are_ watching, not what people would get the most enjoyment out of watching.
Just to be clear/nitpicky, you're talking about an algorithm that starts with the catalog Netflix could afford to license at some moment in time, and then maximizes the viewer's enjoyment for that limited set of content.
That's probably a vastly different level of enjoyment than starting with, say, the catalog of available content listed in IMDB and maximizing for the same thing.
Sure. But they are still at best delivering "the most enjoyable experience where the enjoyment is bounded by the content that Netflix can afford to license."
A user can hit those constraints when they search for something that exists but is not available on Netflix.
That point becomes noteworthy when that same content (or portions of it) are available elsewhere on the net. In my experience this is typically through:
a) downloading a torrent of the content that is missing from Netflix
b) watching all or some portion of that content on Youtube
Just to be clear/nitpicky, you're talking about an algorithm that starts with the catalog Netflix could afford to license at some moment in time, and then maximizes the viewer's enjoyment for that limited set of content.
That's probably a vastly different level of enjoyment than starting with, say, the catalog of available content listed in IMDB and maximizing for the same thing.