Paul argues testing will be pushed aside by measurement of actual performance. He may be right but existing methods of testing are amazingly unsophisticated and

Paul argues testing will be pushed aside by measurement of actual performance. He may be right but existing methods of testing are amazingly unsophisticated and open to improvement. For instance: Most work in the real world is done in teams, but schools tend to grade on average team output, which is vulnerable to free riders. Why not rotate individuals through teams and measure independent contributions with some simple statistics? That's what coaches do for sports like rowing. Knowledge in the real world is rarely all or nothing. It matters how confident you are in what you know. Will you risk your company's future on an algorithm? Why not adjust test results by student's own ratings of their confidence in their answers? Then we could see who can really be trusted with their knowledge.