A teacher has multiple goals at any one time in a classroom. She has multiple goals for any one child: cognitive, social, and emotional goals, and goals not just about what is learned here and now, but what will serve as good preparation for future learning and the creation of life-long learners. She has multiple goals for the children as a group in terms of participation, collaboration, and the preparation of future citizens and active members of the public sphere. She wants to produce learners who can integrate science, ethics, problem solving, and a concern for the deeper truths that the humanities have always offered us as "equipment for living." Judging a teacher by reading, math, and science scores on one-off "drop out of the sky" tests from some corporation is absurd and immoral.
Teachers cannot be held accountable for "annual yearly progress" for all groups because they teach different kids each year. Normal variation and regression to the mean ensures that AYP is impossible. Whether a child succeeds has to do with a lot more than what a teacher does, no matter how good she is. Surgeons can fail because a patient is too sick, often with problems other than the one the surgery is meant to address. Teachers can fail because a child has other problems, as well. Basing accountability on success measured in tests scores and annual yearly progress ensures that teachers will want to teach only those "healthy" children, just as such an accountability system would ensure that surgeons would treat only healthy patients at little risk.
With surgeons, we do not want to see how many people they kill and weed out the bad ones. We want to be sure that the vast majority are competent by the time they get their degrees. We have accountability for medical training and certification. Then we weed out the outliers or surgeons who "turn bad" later, relatively rare cases. We should not just see how many kids teachers damage and weed out the bad ones. We should be sure that the vast majority are competent by the time they get their certification. We should have accountability for certification. And we all very well know the only way to do this is to radically change the School of Education as we know it, a broken institution if there ever was one.
There is another thing we should have accountability for and that is curricula and pedagogies. Here we should let a thousand flowers bloom in test beds and correlate each of them with a rigorous sampled assessment that we all trust, like NAEP or PISA, or some other one we would design. This assessment would not be given to all students and, thus, could not be taught to or cheated. If we find out that a "liberal" approach to math and a "conservative" one both correlate with high NAEP scores, then so be it. There are also different drugs that work well for the same thing, sometimes for different sorts of people. Teachers that can be resourced and developed to become experts in what works for them, their children, and their schools. Then they can stick with it, get better at it, and succeed without having to adopt ever new fashions and fads. I first heard this idea from John Katzman, the man who founded Princeton Review (or, at least, this is what I took away from what he told me).
So, let's have harsh accountability for certification and curricula/pedagogies and rich development for teachers.