That said, we don’t trust systems to understand what “best” means.
Just like Tay can’t tell what not to model its responses upon, no algorithm I’ve seen could come close to the subtlety of actual human interactions.
As an example, sometimes there are friends or family who are fairly passive with a social network, but whose activity—which a bot would interpret as uninteresting—is our way of keeping tuned into their level of depression. Sometimes we act on their messages but often just seeing them is enough (particularly when our primary social activity with them happens outside the feed).
Often there are inside jokes a bot wouldn’t get.
Algorithmic feeds which use activity level on a post/tweet/image are inherently biased against quieter relationships and smaller networks.
I follow high-signal folks like Anil Dash and very low activity folks who are important to me in the same stream. Algorithmic feeds don’t get the subtle differences and fail to put those folks on an even footing.
So, no, no matter how nice the folks are and how best damn product what they’re making is supposed to be, I will continue to reject algorithmic feeds and instead tune my follow activity to just what I can handle.