> You learn at the same university that gives you the degree.
In the United States, this certification happens at the institutional level. Universities are either accredited by either a regional body (good) or a national body (not so good). The accreditation process requires demonstrating that you have competent faculty, that these faculty have designed a reasonable curriculum from each of your degree programs, that the courses taught as part of that curriculum cover appropriate material, and that students are at least nominally assessed.
In addition to accreditation, there's also a reputation effect among employers. We know which of the universities in our region have strong CS programs, and which ones are not so strong. And there's a reason that employers prefer CS students from places like CMU/MIT/Stanford/Berkley/UIUC/etc. The names aren't just names -- those institutions have exceptional CS programs. I'm sure a student slips through the cracks here or there, but the median student from those institutions is almost certainly going to at least have a good background in CS.
> But we also know for a fact that having them decoupled can cause serious problems as well. For one you end up with brain dumps that study for the cert and not for the subject matter. We see this for MCSE as much as the SAT.
Frankly, I think the decouple-and-test model is substantially worse than the accreditation model. Have you seen the contents of the Computer Science subject area GRE? It's from the 1980s. The AP CS exam and model curricula aren't that much better -- definitely not what I would expect out of a CS 1 student at any decent university. In fact, I can't think of a single formal exam in CS that's actually a decent signal of even basic CS knowledge, let alone engineering skill.
> In the United States, this certification happens at the institutional level. Universities are either accredited by either a regional body (good) or a national body (not so good). The accreditation process requires demonstrating that you have competent faculty, that these faculty have designed a reasonable curriculum from each of your degree programs, that the courses taught as part of that curriculum cover appropriate material, and that students are at least nominally assessed.
Yes, I'm aware of accreditation and the process. I've served on a department advisory board for about 5 years now and have seen the process via ABET first hand. I hate to disappoint you, but these accreditation agencies are box-checkers and they don't do too much other than publish some guidelines and look at course material. The university could lose the accreditation and teach the same stuff and nothing would really change for students - in a vacuum student outcomes would be exactly the same except that as you mention...
> there's also a reputation effect among employers.
This has not much to do with accreditation. CMU, MIT, Stanford, etc... could lose their accreditation and nothing would change for the university. Their reputation is far stronger than the governing body. The exceptional programs that are there exist without consideration to ABET. It's just not meaningful. If they said "we are no longer complying with ABET" (at least for engineering) it would discredit ABET and not the university. Not all universities transcend this, but I'd say most brand name ones do (Ohio State, Michigan, Arizona, Georgia, etc.). I'd bet a survey of employers and students would show very little knowledge of the existence of many of these accreditation bodies. It's just a battle not worth fighting professors over, and the accreditation bodies don't look too hard because why kill the golden goose? Everybody just winks at each other. I guess think about accreditation bodies as insurance.
Anyway, that's just my take on these accreditation bodies. I'm sure some are better and some are worse.
But my main point was that education itself is decoupled from certification. The employers (as you mentioned) care about certification, but that is the degree from the prestigious university, not whether ABET says that the curriculum meets some standard. You can learn computer science or any other topic to a high degree of proficiency, but employers still care about the certification. So you might as well just get the certification if you care about education as a means to a job.
AFAIK ABET's process is different from regional accreditation, probably in part because ABET's historical roots were in Engineering where there is an explicit assumption of eventual external certification (the PE exam).
> The university could lose the accreditation and teach the same stuff and nothing would really change for students - in a vacuum student outcomes would be exactly the same except that as you mention...
This is definitely not true of regional accreditation -- a degree from an unaccredited institution isn't worth the paper it's printed on in many fields. Also, as you point out, it's a low bar. If a university is accredited it may or may not do a good job at education/assessment. But a university that loses regional accreditation is almost certainly doing a horrible job at both.
ABET is a bit different -- lots of quality programs aren't ABET accredited. But still, in traditional Engineering fields, losing ABET accreditation can also have a material effect on students' outcomes (can you even sit for the PE exam without an ABET accredited degree?)
> these accreditation agencies are box-checkers and they don't do too much other than publish some guidelines and look at course material.
Yes, there's no silver bullet and the map is never the territory. The only alternative I'm aware of is standardized exams, which, well, look at US K12 or the CS GRE.
>> there's also a reputation effect among employers.
> This has not much to do with accreditation.
Right, they are two separate things but both effect institutional behavior.
> But my main point was that education itself is decoupled from certification. The employers (as you mentioned) care about certification, but that is the degree from the prestigious university, not whether ABET says that the curriculum meets some standard. You can learn computer science or any other topic to a high degree of proficiency, but employers still care about the certification. So you might as well just get the certification if you care about education as a means to a job.
In the United States, this certification happens at the institutional level. Universities are either accredited by either a regional body (good) or a national body (not so good). The accreditation process requires demonstrating that you have competent faculty, that these faculty have designed a reasonable curriculum from each of your degree programs, that the courses taught as part of that curriculum cover appropriate material, and that students are at least nominally assessed.
In addition to accreditation, there's also a reputation effect among employers. We know which of the universities in our region have strong CS programs, and which ones are not so strong. And there's a reason that employers prefer CS students from places like CMU/MIT/Stanford/Berkley/UIUC/etc. The names aren't just names -- those institutions have exceptional CS programs. I'm sure a student slips through the cracks here or there, but the median student from those institutions is almost certainly going to at least have a good background in CS.
> But we also know for a fact that having them decoupled can cause serious problems as well. For one you end up with brain dumps that study for the cert and not for the subject matter. We see this for MCSE as much as the SAT.
Frankly, I think the decouple-and-test model is substantially worse than the accreditation model. Have you seen the contents of the Computer Science subject area GRE? It's from the 1980s. The AP CS exam and model curricula aren't that much better -- definitely not what I would expect out of a CS 1 student at any decent university. In fact, I can't think of a single formal exam in CS that's actually a decent signal of even basic CS knowledge, let alone engineering skill.