Just from the implementation side of things, you have a few problems: - The majority of people will never seek out settings like this, so the engineering effor

Just from the implementation side of things, you have a few problems: - The majority of people will never seek out settings like this, so the engineering effort you spend on these sorts of classifiers is only going to benefit a minority. Facebook has a lot of engineers to throw at these problems, so that's not the end of the world, but all else equal they'll prefer improvements that have a better chance of helping the whole userbase. - Trying to coax people to look at settings is tricky. Users have very limited patience for interruptions, and a lot of it is already taken up with important flows like "hey did you know your password is terrible?" - Exposing settings like this kind of ties down some of the details of your system. Say you start with 6 different categories, and I choose 3 of them. Fast forward a couple of years, and now there are 30 categories. If I haven't changed anything, do I still have only 3, or do I now have 15? That's gonna depend on the details, of course, but one problem is that not everyone will interpret the old-categories-new-categories relationship in the same way, and some users are going to end up confused or disappointed. - Users don't necessarily know what they want, especially looking farther into the future. Maybe it's an election year, and I get tired of politics, and I say "no political news for me please." A year later, maybe I'd be interested in some of the political news that's happening, but there's nothing prompting me to go back and change that setting. Trying to put a time-limit on these filters, or prompting users to change them, adds more and more to the clutter.