The fact that I lingered longer on a post doesn't indicate the polarity of my interest, just the magnitude at best.

Knowing that you linger longer on that topic is good enough to know that the topic is likely to get further engagement from you though - people like to spend time arguing about things that annoy them as well so you don't only want to show things that make people smile.

If you interact in any way with a post, you can do sentiment analysis on the text too to figure out their thoughts on the topic.

I just don't see facial analysis being that powerful over the above. You're already giving Facebook and Twitter a ton of information about what you enjoy via likes, shares and follows, without having to get into the subtleties of what it means if you half-smile or frown when you see a post.

Seriously why is the discussion anything but delete all tech immediately and bomb silicon valley?

Too late, the ML just flagged me as being anti-ML. I'll probably be assassinated soon.

The problem is not the tech or the ML. It is the people using and creating it.