Abstract:

Recent advances in Educational technology changed drastically the way we teach and learn and leads to new educational approaches overcoming the traditional obstacles of time and space. Indeed, one of the ongoing issues regarding today’s educational technology is supporting learners in the way they learn. For instance, learners have different backgrounds, knowledge and goals; they proceed according to their own learning style and might be interested in different contents at different stages of the learning process.

In this contribution, we focus on identifying learning style: a set of cognitive constructs which is about how do learners behave and react in a specific learning environment. In the literature, two distinct approaches are discussed and could be classified into implicit and explicit ones. The explicit approaches determine the learning style by using a questionnaire to fill. However, those questionnaires lead often to misunderstanding and misidentifying the learning style because learners are not always aware of their own learning style. In the other hand, implicit approaches propose to analyze learner’s traces to detect their learning style. They use intelligent algorithms that rely on machine learning techniques to generate relevant information about the learner. The results are more effective than the traditional methods of identifying the learning style but it has notable drawbacks: at the beginning of online learning activity, the platform holds few information, a common problem known as e-learning cold start problem. Therefore, those algorithms need existing data to generate efficient results.

To this end, this paper proposes a new approach to identify learning styles using Computerized Adaptive Testing: an adaptive questionnaire that collects information about learner knowledge level and learning style. Still, one question remains to be answered: how could Computerized Adaptive Testing identify learning style?

One solution to the above problem is to use Item Response Theory (IRT) for an educational purpose. It’s a theoretical framework used for designing Computerized Adaptive Testing. The IRT relies on psychometric models that assumes that it is possible to estimate an individual latent trait based on his/her answers to a set of tailored questions. To design adaptive tests, we suggest adapting the IRT to our application context. The learner responses during a test are considered as a stochastic process in which the probability of giving a correct answer or not, depends on previous asked questions/answers and relies on learners themselves, such as their cognitive and metacognitive resources, affective and social components. While answering an item, the learner is showing an underlying ability that the test is trying to capture. In our case, we are trying to capture information about student learning style based on his/her behavior while answering questions.

To apply the IRT concepts to identify learning style we proceed as follows:
1. Build a questions bank. These questions have the particularity of being presented in distinct formats, in other words, the same question can be asked differently to different learners.
2. Use Item Response Theory to design adaptive tests: provide learners with questions aligned with their potential learning style during the test and analyze their behavior.

Keywords:

Learning Style, Educational Technology, Personalized learning, Computerized Adaptive Testing, Learner Profile.