The teaching of Informatics in secondary education constitutes a distinct field, as it combines a constantly evolving subject area, laboratory and theoretical requirements, elements of digital culture and social dimensions, cognitive difficulties in understanding core concepts, and at the same time the need for creativity, collaboration, and interdisciplinary approaches. More specifically, with regard to learning programming, the typical teaching approach is usually based on the use of high-level programming languages and a programming environment or text editor in order to write and execute programs for solving simple computational problems. Nowadays, however, an alternative approach for introducing lower secondary school students to programming is educational robotics, which connects programming with the physical world and offers a more experiential, engaging, and collaborative learning environment. As research on Informatics teaching techniques continues to evolve, the reasonable question also arises of how these techniques can be evaluated so that their weaknesses can be identified and improved. Traditionally, the use of student questionnaires constitutes a reliable way of assessing the teaching process, as it allows the teacher to investigate the extent to which the learning objectives were achieved and how students experienced the lesson. More recently, machine learning techniques from the field of artificial intelligence offer new evaluation possibilities, which can be applied early in the school year in order to identify weaknesses both in the teaching technique and in students’ learning progress, thus allowing timely corrective interventions. Within this context, the present research approaches the design and evaluation of teaching as a unified process, aiming at a more comprehensive way of assessing teaching techniques through the combination of traditional and contemporary methods. The present research focuses on the design, organization, implementation, and evaluation of an alternative teaching scenario for learning programming by lower secondary school students, based on educational robotics, the flipped classroom model, and the use of an asynchronous distance learning environment under conditions of limited weekly instructional time. This teaching scenario aims to transfer part of the learning process outside the classroom, so that students can access the educational material independently of time and place and work through it at their own pace, while face-to-face instructional time is used mainly for collaboration, inquiry, experimentation, construction, and programming of robotic systems. Within this framework, a variation of the Jigsaw technique was also developed, specifically adapted to the requirements of educational robotics, with the aim of organizing student collaboration in groups more effectively and supporting their active participation in problem solving. This teaching intervention was initially evaluated by the students themselves through a questionnaire, in order to investigate the degree to which this specific learning environment responded to their needs for learning programming. The results showed that students responded positively to the process of learning programming through educational robotics, that the possibility of seeing the results of their programs through the movement of a robot enhanced both their understanding and their interest, and that the organization of the lesson within a flipped classroom environment supported their access to the material, study at their own pace, communication, and collaboration. At the same time, it was found that content formats such as examples with images and videos significantly helped students understand programming more effectively. The research then proceeded to make use of data from the learning process in order to develop a machine learning model aimed at predicting students’ final performance in the Informatics course. Although such techniques have been studied mainly in post-secondary education, where large datasets are usually available, in the present research they were applied at the secondary education level despite the limitations imposed by the small size of the dataset. Students’ initial grades from quizzes and assignments, together with demographic data, were used in order to build and test different predictive models. The results showed that, even within this context, it is possible to develop a sufficiently accurate model for predicting student performance. The CHAID algorithm demonstrated the best overall performance in terms of accuracy (82.14%) and AUC value, compared with other algorithms tested, such as QUEST, C5, BN, and RF. At the same time, the most important predictor of final performance was identified, offering useful indications of which elements of the educational material and learning activities appear to have the greatest instructional value. Overall, the present research highlights that learning programming in lower secondary education can be effectively supported through an alternative, student-centered, and technologically enriched teaching framework that combines educational robotics, the flipped classroom, collaborative learning, and the use of data derived from the learning process itself. At the same time, it shows that the combination of questionnaires and machine learning techniques can offer a more comprehensive way of evaluating teaching techniques, strengthening both the improvement of teaching and the timely support of students within everyday school practice.
Δηλώνω ότι έλαβα γνώση και ανεπιφύλακτα συμφωνώ και αποδέχομαι τους
περισσότερα
