Dear dmund

It"s a good idea, CAT is a useful approach to detect informations about both test and respondents, you might need to do your own algorithms to do such, I dont have experience,

IRT  models used widely in CAT they aproved  good results but i would tell you about nonparametric item response theory models, they are suitable for  ordinal data also CAT developed to be applicable with them. hope that these will help you

best wishes

http://echo.edres.org:8080/scripts/cat/catdemo.htm

http://www.google.com.sa/url?sa=t&rct=j&q=&esrc=s&frm=1&source=web&cd=15&cad=rja&uact=8&ved=0CEAQFjAEOAo&url=http%3A%2F%2Fwww.lsac.org%2Fdocs%2Fdefault-source%2Fresearch-%2528lsac-resources%2529%2Fct-04-05.pdf&ei=p2RWVaCiLcvZU42agcgB&usg=AFQjCNFLOVU36gegoFGLZ6__lK5PLF0OJw

http://www.google.com.sa/url?sa=t&rct=j&q=&esrc=s&frm=1&source=web&cd=20&cad=rja&uact=8&ved=0CGoQFjAJOAo&url=http%3A%2F%2Fwww.professeurs.polymtl.ca%2Fmichel.desmarais%2FPublications-Michel%2Fdesmarais-pu.pdf&ei=p2RWVaCiLcvZU42agcgB&usg=AFQjCNHE9tTy05vb7D_UpiH8afCb7-ighA

The concerto platform

http://code.google.com/p/concerto-platform/

is a very good device for online adaptive testing!

Hello Edmund,

Concerto is a nice platform to work with. At the same time it will depend of you programming skills. If I not mistake, you have to pay depending on the number of subjects and to have access to other things.

You always can programmer in R, which is the same source code as Concerto.

There are two companies, that I am aware that work with CAT. One is CITO (Netherlands) and the other is ASC (USA). The latter is really good and very friendly. I didn't have any contact with CITO.

IACAT as suggested by Aries is a good place to start. There are a lot of articles available to download. 

Usually, the studies are divided in the CAT methodology and applied. I'm not sure with which one are you interested in. If you are interested in automatic generation of item you may have a look for Heinz Holling. He has a very nice paper regarding automatic generation of item in an intelligence test.

You should examine the PersonFit application. To access the application, you could contact Komi Sodoke .

You could also refer to David Magis and Professors Gilles Raîche  and Sébastien Béland

https://www.researchgate.net/publication/221423223_The_adaptive_and_intelligent_testing_framework_PersonFit

Article IMS-QTI sub-standards in computerised adaptive testing and interfacing

Conference Paper The adaptive and intelligent testing framework: PersonFit

Article La plateforme d'évaluation adaptative des apprentissages: PersonFit

Conference Paper Présentation d’un outil logiciel pour soutenir la détection ...

It is impossible to develop a valid CAT without building it completely around IRT (http://pareonline.net/pdf/v16n1.pdf).  Concerto is certainly a possible platform for you to use because it is extremely inexpensive, but there are issues with using it, as Dario noted.  It is primarily designed for research use.  If you need large-scale use, I'd recommend looking at ASC (also noted by Dario) at www.assess.com, who also provides IRT calibration software, Xcalibre. (Disclosure: I work there)  DeBeer's work was built on previous versions of our platform.

The best general resource on CAT is www.iacat.org, as mentioned above by Yulianto.

Hello Edmund,

In case you are looking for IRT based adaptive testing R packages , please go through catR , mirt and mirtCAT packages. Especially mirtCAT package , it provides tools to generate an HTML interface for creating adaptive testing.

Hope it helps you.

You should develop software for your own, not too difficult to deploy in R or Python or so on.

mirtCat is fantastic and easy to understand with a wide range of stop, item selection, and constraints. It's also hosted on RShiny making it free and easily distributable (https://cran.r-project.org/web/packages/mirtCAT/index.html)