Over the last decades language assessment testing has been done through the use of computers, and starting from '90s we hear of Computer-Based Testing (CBT) and Computer Adaptive Testing (CAT). CBT and CAT have significantly impacted assessment methods, especially in terms of large-scale group evaluation, because of their significant advantages over traditional paper-and-pencil tests.
As mentioned by Chalhoub–Deville et al. (1999) CBT has remarkable potential benefits such as individual administration of tests, reduced pressure in terms of scheduling and supervising tests, more frequent and convenient testing; greater standardization of test,
collection and storage of various types of information about test takers responses (e.g., response time, item review strategies, items omitted or not reached, etc.),· possibility of special accommodations for test takers with disabilities, etc.
The language tests currently used by The University of Applied Sciences and Arts of Southern Switzerland (SUPSI) are classical CBT tests, thus making all these potential
advantages possible. CAT would be a further development, improving the accuracy of the evaluation, as it is based on the delivery of questions based on the student’s previous answers, and on a continuous tuning of the level of difficulty of the questions.
As highlighted by Green (1983), CAT has many additional benefits such as improved test security, by providing large banks of questions, making it extremely difficult for test
takers to learn the answers to all of the questions; rate of response can be used as an additional way of assessing the ability of a test taker; all test takers are challenged, and none are discouraged, as most questions are near their ability level, etc.
However, the successful implementation of CAT has been limited by a number of technical and practical factors. Ockey (2009), pointed out some interesting controversial issues in the use of CAT. In particular, there is general disagreement on how Item Response Theory (IRT, i.e., the main paradigm for tests design) can be implemented within CAT, and regarding the most appropriate scoring algorithm.

 Artificial intelligence applied to area of decision support system (DSS), seems to be a suitable framework for the development of CAT tools. A number of based either on expert knowledge to be quantified by proper elicitation techniques, of a statistical processing, have been proposed during the last ten years. Among others, Vomlel (2004) emphasized the importance of a proper modeling of the answers to the questions and the skills to be mastered by students.

We plan to use these probabilistic strategies to implement DSS in CAT field. Different strategies will be applyed on the basis of the specific aspects that concern language skills evaluation, and will be identified by languages teachers at the SUPSI Languages centre.
The main goal is to evaluate the ability of the students in a more accurate and purely automatic manner, implementing specific data-mining techniques related to survival analysis for a better profiling.

Chalhoub–Deville, M., & Deville, C. (1999). Computer adaptive testing in second language contexts. Annual Review of Applied Linguistics, 19, 273-299. Green, B. (1983). The promise of tailored tests. In H. Wainer & S. Messick (Eds.), Principals of modern psychological measurement (pp. 69–80). Hillsdale, NJ: Erlbaum. Ockey, G. J. (2009). Developments and Challenges in the Use of Computer Based Testing for Assessing Second Language Ability. The Modern Language Journal, 93(s1), 836-847. Vomlel, J. (2004). Building Adaptive Tests using Bayesian networks. Kybernetika, 40 (3), 333-348.