On Thu, Mar 31, 2005 at 12:34:24AM +0200, David TAILLANDIER wrote: > > >>What about an excrutiatingly large question pool designed for adaptive >>testing? Integration of free-response/essay questions? A live-distro >>that requires actual skill to accomplish a real task that can be >>recorded and graded or automatically analyzed? The key I would think >>would be to make the testing as multi-faceted as possible with >>emphasis on skill and knowledge. > >A large pool, why not, but who is able to do it ? The problem is first >to create questions, then to validate and rate them : every question >does not have the same 'power'. >An excrutiatingly large pool, well, it just may be impossible.
A lot of the topics discussed recently have been well hashed before, check the archives http://lists.nycbug.org/pipermail/bsdcert/ google probably knows. ...more inline... On Sun, Dec 19, 2004 at 12:35:29PM -0500, George Georgalis wrote: >On Sun, Dec 19, 2004 at 07:52:06AM -0500, Sunny Dubey wrote: > >>Finally, if you establish a cert there is no real viable market for, >>I don't think you'll ever beat the chicken and the egg problem. >>(Managers look for people with those certs, and people getting those >>certs because they think mangers are looking for them.) > > >I think the key is to develop a self substantiating cert, where the >value is not from reputation but from content. For example if 1,000 >difficult, 1,000 med, and a 1,000 easy problems are "ported" to each >OS; then, candidates choose an OS and tackle 10 random difficult, 20 >med and 30 easy problems in 2 hours. > >The cert has no reputation but anyone can see the questions (possibly >in a restricted way) and determine the value of completing the test for >themself. > >There should probably be other aspects of a cert, but this test >will show that the candidates can just fix the problem (no time for >research). Even if they study all the questions, completing those 40 >random problems in 2 hours would demonstrate the competence the test is >certifying. (adjust numbers to suit) > >The real hotshots can win a cert for each offered OS, with no failures >each year/version it is offered. I think remote testing should be an option even if it does not satisfy a full hardware cert. Thinking about validating and cheating... cert for a pgp sig could work. It would still be up to the employer to determine if the candidate was honest, ie no assistance on taking the remote exam. An online test with a large comprehensive question base described above, could be utilized by an employer (or candidates) to validate competency. If the point wasn't obvious: a categorized comprehensive (or as near so as possible) question set removes the issue of candidates memorizing the 100 question/answers while not knowing anything. If they memorize 10000 question/answers and they still don't know anything, there must be personality issues. A job of the board and community would be to submit and audit question/answers. Not saying this should be the exclusive test, and the method wouldn't work until there is a large question base, but I think it could go a long way to the evaluation. // George -- George Georgalis, systems architect, administrator Linux BSD IXOYE http://galis.org/george/ cell:646-331-2027 mailto:[EMAIL PROTECTED] _______________________________________________ BSDCert mailing list [email protected] http://lists.nycbug.org/mailman/listinfo/bsdcert
