Different categories of online assessments are used by Learning Analytics approaches to monitor students’ progresses. Questionnaires are a valid tool to assess students’ performance. In this perspective, it is important to investigate the relationship between questionnaire answers and students’ skills. Classical Test Theory (CTT) and Item Response Theory (IRT) study the relationships between questionnaires’ answers and hidden latent concepts leading at those results. However, several studies have proven the superiority of IRT in respect to CTT [1]. In the educational domain, the aim of IRT is to provide a mathematical model to predict and evaluate students’ abilities that are measured through a questionnaire. The main idea behind this theory is that latent abilities underlie both the students’ performances and the test items. The term “ability” embeds different cognitive students’ skills, that are strictly related to the topic under evaluation (e.g. solving a mathematical problem requires different skills compared to text understanding). Understanding students’ skills or the lack of these skills can be used as feedback for supporting students to identify their learning needs, to measure teaching effectiveness and discover difficult topics [2]. However, latent skills are difficult to be manually extracted since they are topic specific, thus requiring expert analysis. Non-negative matrix factorizations (NMFs) are dimensionality reduction techniques that are able to describe original data as an additive linear combination of hidden factors [3]. In the educational domain they have been proven to be effective for extracting hidden skills from questionnaires items responses, and for profiling students in terms of these skills [4]. NMFs are highly interpretable since hidden skills are described in the same space of the original data, thus helping an intelligent analysis of the results from domain experts. In this work, students’ answers to the Maths Challenge competition, that has been carried out in 2021 at the University of Foggia, have been analyzed through NMF. Moreover, results have been explained in terms of a-priori knowledge on item’s topics, and mathematical assessment measures. The article is structured as follows: data and the use of NMF algorithms in educational domain will be briefly described in paragraph 2. Results will be summarized in paragraph 3 and conclusions and future works will be detailed in paragraph 4.
Intelligent Knowledge understanding from students questionnaires
Luca Grilli;
2021-01-01
Abstract
Different categories of online assessments are used by Learning Analytics approaches to monitor students’ progresses. Questionnaires are a valid tool to assess students’ performance. In this perspective, it is important to investigate the relationship between questionnaire answers and students’ skills. Classical Test Theory (CTT) and Item Response Theory (IRT) study the relationships between questionnaires’ answers and hidden latent concepts leading at those results. However, several studies have proven the superiority of IRT in respect to CTT [1]. In the educational domain, the aim of IRT is to provide a mathematical model to predict and evaluate students’ abilities that are measured through a questionnaire. The main idea behind this theory is that latent abilities underlie both the students’ performances and the test items. The term “ability” embeds different cognitive students’ skills, that are strictly related to the topic under evaluation (e.g. solving a mathematical problem requires different skills compared to text understanding). Understanding students’ skills or the lack of these skills can be used as feedback for supporting students to identify their learning needs, to measure teaching effectiveness and discover difficult topics [2]. However, latent skills are difficult to be manually extracted since they are topic specific, thus requiring expert analysis. Non-negative matrix factorizations (NMFs) are dimensionality reduction techniques that are able to describe original data as an additive linear combination of hidden factors [3]. In the educational domain they have been proven to be effective for extracting hidden skills from questionnaires items responses, and for profiling students in terms of these skills [4]. NMFs are highly interpretable since hidden skills are described in the same space of the original data, thus helping an intelligent analysis of the results from domain experts. In this work, students’ answers to the Maths Challenge competition, that has been carried out in 2021 at the University of Foggia, have been analyzed through NMF. Moreover, results have been explained in terms of a-priori knowledge on item’s topics, and mathematical assessment measures. The article is structured as follows: data and the use of NMF algorithms in educational domain will be briefly described in paragraph 2. Results will be summarized in paragraph 3 and conclusions and future works will be detailed in paragraph 4.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
