How discredited is VAM actually? (This is a rathole, and we can probably table it, but I don't come to this discussion assuming radical changes to teacher evaluation are a terrible idea, the way I do with broken-windows policing). It clearly predates Outliers by a lot.
Diane Ravitch has an entire book about this. If you read the two chapters about the epistemology of the statistics behind VAM, you can see it's actually pretty similar to Theranos, an idea that can't even work on principle regardless of the technology or how it's implemented:
One on-the-ground outcome of these practices I've heard about:
You're a math teacher. Like half the incoming class is horribly deficient at one part or another of math that was last heavily-covered several years earlier. I'm talking, you're a 7th grade math teacher, and some of these kids aren't able to do long division, plus probably a bunch of other stuff (it's rarely just one thing). This can happen because a single teacher a single year was kinda shit and no-one since has tackled this problem.
You have a choice.
1) The right thing to do is ignore practically everything about what you're supposed to teach them in your grade—not for all the students, but for those with huge gaps in knowledge—and try to fix some of those problems, until all the worst problems are addressed, and only then (and you probably won't have time) try to teach them the new content. They are all but certainly doomed to never truly improve at math if you don't do this.
This will result in those students bombing the standardized tests on which you'll be judged, and on which your career hinges. Good chance they'll do worse than they did the year before. It's also a bunch of extra work because you're coming up with and applying way more lesson plans, activities, assignments, et c. But, they'll in fact be better off.
OR...
2) You can mostly ignore those problems and teach them as much as you can of this year's content. You're a good teacher (else you'd not even consider the first option) so you can do a decent-enough job of getting them to the point they can at least pattern match and guess their way to... well, not success, but something. Yes, their scores will still be bad, but it might mean they get 35% right instead of 15% (if you'd taken the first option), and maybe last year they only got 32% right, so it still looks like you did a good job (the tested content's different each year, so even a small % improvement in scores is good)
Almost nobody chooses the first option because they'd have admin up their ass in a hurry. If they're good and really dedicated they might put in a ton of extra effort and their own time to try to put together supplementary instruction programs of one sort or another. This will likely gain them no extra compensation, and probably involve a whole lot of admin- and politics-induced bullshit. It might even make some enemies. No fun.
Repeat for a few years until the way-above-average teacher burns out and quits to go get paid & treated better in another career, with way less stress.
She was one of the chief architects of NCLB, so on that basis you'd expect her to support high stakes testing and VAM. But obviously she's completely changed her opinion since the Bush era, largely based on the academic research that has been done since on the effects and effectiveness of those policies.
