Abstract:

Identifying groups of students that share similar intelligence characteristics can help teachers and instructors better understand the preferences of their learners and to facilitate group work in the classroom, so that teachers can provide more personalized learning activities that match their intelligence types. According to Howard Gardner, the theory of multiple intelligences (TMI) states that humans possess eight intelligences, in varying degrees (musical intelligence, special intelligence, logic-mathematical intelligence, ...).

The purpose of this paper is to answer to research question: How can be improved the way in which a teacher can form groups and how to choose adaptive learning activities for each group of students?

The objective of this research is to form groups of students based on their intelligence types according to this theory (TMI). For this purpose, clustering methods are used, which are unsupervised learning techniques that allow similar entities to be grouped together based on their characteristics. This study reviews clustering methods, evaluates and discusses the performance of two clustering algorithms (the hierarchical clustering algorithm and the k-means algorithm) on a practical multiple intelligence dataset, in order to compare the results and determine which algorithm is better at classifying students into clusters (groups of students) based on their intelligence types.

The data was collected using the multiple intelligences questionnaire that included 80 questions, each of which referred to one of the multiple intelligences according to Gardner's theory. And the students are asked to grade each sentence between one and five to indicate how student is satisfied (strongly disagree to strongly agree). The sum of the scores in each category (10 questions in each category) determines the degree of each type of intelligence (multiple intelligence data).
The research was conducted with a case study of a Moroccan high school to teach programming to high school students. And as a result, we found 8 clusters using K-mean algorithm and 9 clusters using CAH algorithm. These results are applied to produce adaptive and personalized learning activities for each cluster (groups) with a high similarity of intelligence type and the result is compared.

This research has led us to find that these clustering techniques show comparable results that can be useful to teachers for work in groups in such a way as to tailor classroom learning activities to students' intelligence types and to help students maximize their strengths and address their gaps.

Keywords:

Clustering, learning, multiple intelligence, group formation.