Comment by ender7 | Hacker News Reader

Comment by ender7 | Hacker News Reader

Here is the crux of the matter:

In the book, Mr. Khan also advocates for a separation of universities’ teaching and credentialing roles, arguing that if students could take internationally recognized assessments to prove themselves, the playing field would be leveled between students pursuing different forms of higher education.

Removing our heavy reliance on higher ed credentialling seems central to his idea, and yet the proposed solution is simply "internationally recognized assessments", which sounds suspiciously like "standardized tests" to me.

Sadly, many professions are not well-evaluated by standardized testing. This is not for lack of trying -- we already have the GRT and all of its related subject tests, plus the entire raft of tech-related certifications that aren't worth the paper they're printed on (remember "A+ certification"?). Grad schools look at GRT results, but usually only to make sure you didn't completely bomb the test.

This is the hard part, and will require a lot of careful though to untangle.

Standardized tests don't have to be multiple-choice tests from a question bank, though. You can have practical sections, like in the CCIE. You can have long-form scenarios you have to analyze like on part 3 of the CFA. You can have the exams only given on specific dates at specific times of the years, as is done on several exams, so that the questions are always secret and not reused.

The highest quality certifications comnbine more than one of the above techniques. For instance, you might make phase 1 be just a multiple choice test. This mainly acts as a filter so that not too many resources are wasted scoring the more comprehensive sections for test takers who aren't really serious. Then phase 2 testers might take a more open-ended test with essays, etc. Finally, people who pass phase 2 might have some kind of practical, similar to medical boards or the network building in the CCIE.

This kind of process might not be cheap, either. A lot of people's time are involved in creating, administering and grading of this kind of rigorous testing process, but it even if it costs $2000 - $3000, it is still much cheaper than a university. It is still much more egalitarian as it is not open only to the chosen few.

Grad schools look at GRT results, but usually only to make sure you didn't completely bomb the test.

You will not get into a top ten Economics grad programme without a perfect GRE Math score unless you are really impressive on some other metric. Neither (near) perfect grades nor undergraduate research of a high calibre will either. These are basically expected. To my knowledge Math and Physics have equivalent standards. These programmes (Top 10) graduate the majority of Ph.D.s in their fields. Within them it looks like GRE doesn't matter but that's because everyone there reaches a high minimum.

But GRE scores predict research productivity so they're definitely measuring something meaningful. Programmes that are less selective have worse outcomes. They're less rigourous than programmes that can assume greater ability and preparation of their students.

Many people love to hate on standardised tests but IQ tests and IQ test equivalents like the GRE, GMAT, LSAT, SAT etc. are the best predictor of workplace performance besides work sample tests. For the time investment involved they are absurdly effective. If you want to improve the usefulness of certifications like A+ you could just show percentile score rather than a simple Pass/Fail.

I regret that being on my phone renders me unable to provide citations but if you HNsearch tokenadult worksample tests you'll come up with sone relevant stuff.

> Sadly, many professions are not well-evaluated by standardized testing.

1) This isn't really true. Bar exams, and medical and accounting equivalents do quite well.

2) It's just as arguable that many professions are not well-evaluated by university accreditation.

I wouldn't be so quick to dismiss certifications. Having an A+ cert at least reasonably signifies someone could format a hard drive and install ram.

These tests do not have to be on paper alone, either. Look at the CCIE from Cisco. That has an extremely difficult lab component that no amount of "teaching to the test" is going to help you overcome (which is why having one pretty much guarantees a six figure income).

There are legitimate ways of doing standardized testing. I suspect the professions of engineering and law can be useful guideposts. There have been discussions before on how engineering as a whole is like or unlike software engineering... I'll leave that bone of contention alone.

But I don't think Khan is suggesting that every profession be so assessed: just that, if an assessment is merited, just put it out there as available rather than required.

Reply on news.ycombinator.com