Take a typical thermodynamics course. Right now, MIT teaches a full term thermo course for mechanical engineers. The exams, including the final exam, cannot possibly test all aspects of thermo, so they sample, but they sample knowing that the students have been exposed to (via p-sets and lecture) the parts of thermo not directly tested.
Now, decouple the exam part and have someone else issue the same exact exam as the MIT profs give, but without requiring the lecture or p-set component. Now, I can imagine there would be a profitable industry player, call them the "Minimum Instructions for the Test", who might teach you only the three laws, enough math to work out a Carnot cycle problem, and the background for one energy balance problem, in other words, just enough to pass the certification exam.
The students from Massachusetts Institute of Technology will have passed the same test as the students from Minimum Instructions for the Test and yet I can have confidence in the former's ability to solve a basic on-the-job thermodynamics problem and not in the latter.
IOW, I think it is, in effect, creating a new problem by eliminating the lecture, office hours, section meetings, and problem sets aspect of the education process and focusing only on the final exam aspect. I'm really not trying to erect and eviscerate a strawman here; apologies if it seems that way, but I'm trying in this post to be very concrete about the problem I see with test-only certifications.
You are very clear. To put in abstract terms: you argue that the education process carry some value which is difficult (or even impossible) to test, making it inherently more valuable than the certification process. I can't say I completely agree with the argument, but I can understand it.
I still think that the tests, if elaborated by entities fully dedicated to them, could grow to be way more advanced than they are today. But even in the case where this growth does not happen, having separate entities dedicated to perform test-only certification could be beneficial.
By the way, if you still read this: suppose you have student A and student B. Student A attended to all classes, did exercises etc. but failed the tests. Student B missed all classes, but passed every test. Do you think student A should be certified? What about student B? Who would you hire?
Whether student B should be depends on the quality and thoroughness of the test, but assuming that is high, then yes.
Whether I'd hire student B, who has shown a propensity to not show up to classes, depends on why they missed them and whether I think that represents a risk in the work-world. Having someone who can pass the test but can't bother to show up to work doesn't help me very much either... ;)
Now, decouple the exam part and have someone else issue the same exact exam as the MIT profs give, but without requiring the lecture or p-set component. Now, I can imagine there would be a profitable industry player, call them the "Minimum Instructions for the Test", who might teach you only the three laws, enough math to work out a Carnot cycle problem, and the background for one energy balance problem, in other words, just enough to pass the certification exam.
The students from Massachusetts Institute of Technology will have passed the same test as the students from Minimum Instructions for the Test and yet I can have confidence in the former's ability to solve a basic on-the-job thermodynamics problem and not in the latter.
IOW, I think it is, in effect, creating a new problem by eliminating the lecture, office hours, section meetings, and problem sets aspect of the education process and focusing only on the final exam aspect. I'm really not trying to erect and eviscerate a strawman here; apologies if it seems that way, but I'm trying in this post to be very concrete about the problem I see with test-only certifications.