I can't go into details, but in general we closely track analytics for our product (like most companies with websites would) and run AB experiments to see how changes impact user behavior.

I really question drawing a conclusion from analytics saying people are using it versus actually liking it. I know some folks at work have used it and they regard it as broken enough to call in a ticket about their device being broke.

Analytics is a pretty well known idea. I question the base theory that it will answer this specific question.

Also, given that the indicator for an AMP page is a greyscale, hard to see icon and words, people who don't want to use AMP might hit it because they are in a hurry and the text they are looking at is black and noticeable.

And they are experimenting with removing the AMP page indicator in search results.

As in it won't go to an AMP page ever or as in it will go to an AMP page without any indication?

The latter.

I don't think usage implies some sort of favorable opinion. They don't really have a choice. Many people may not even realize what is happening.