Nobody said anything about not having standardised testing. The problem here is high-stakes standardised testing. You can know stuff about a system that you ca

Nobody said anything about not having standardised testing. The problem here is high-stakes standardised testing. You can know stuff about a system that you can't measure completely (i.e. any system) by sampling it - you'll get a lot of noise and even some systematic biases, but as long as you maintain an awareness of that you will know something. The problem comes when you want to control things. If you create a feedback loop by attaching strong incentives to the measures you are using to acquire knowledge, then you end up with neither control nor knowledge. You're no longer taking a representative sample, just measuring the gain of your feedback loop. And if you apply the incentives at a granularity of measurement such that noise overwhelms the signal (by a factor somewhere between 7 and 100, according to the article)...