"Works" to me necessarily includes safe, reliable, and gets the job, including variations on the job, done. As an instructor, of course I see things I don't "like" but that's not the test. Instead, I look for ways to see whether it works for the pilot.I agree, but in the context of works and is SAFE. I've seen lots of pilots develop techniques that "work" for them, but it ignores the larger picture. I think that is probably a reason behind a significant number of GA accidents. Much of my training revolved around what could happen, not what usually happens. Unfortunately, after training, many of our habits form around what usually happens and that can be a setup for failure.
To use your own example, I might well think to myself, "Gee, this pilot doesn't use the AI at all. I think that's a bad idea." The answer isn't "change it." It's let's see how it works in a busy terminal environment with establishing descents and other things. And we'll toss in a couple of reasonable things to go wrong and see if it breaks down easily.
I personally use and teach use of the AI. I even corrected a scan that wasn't working for a pilot because of faulty use of the AI. But, having hung out on this board for a while, I'm willing to bet that guys like EdFred, who says he doesn't even look at the AI any more, don;t have the problem you envision with the technique.
If I sound sensitive on the subject of instructors trying to change things, I am. My first IPC about a year after getting the rating was with a CFII who insisted a technique taught to me in my IR training was "wrong". It wasn't. It was just different from what this instructor believed was the one true way.
But my efforts to change it made the technique I was originally taught and the new "right" one he insisted on completely useless to me.