An analysis of computational thinking development through personalized learning paths of programming challenges in a block-based maze game by measuring learners’ performance and challenge difficulty

• Autores: Ioanna Kanellopoulou
• Directores de la Tesis: Mariluz Guenaga Gómez (dir. tes.) , Pablo Garaizar (dir. tes.)
• Lectura: En la Universidad de Deusto ( España ) en 2022
• Idioma: inglés
• Tribunal Calificador de la Tesis: Arnon Hershkovitz (presid.) , Diego López de Ipiña González de Artaza (secret.) , Cristian Olivares-Rodríguez (voc.)
• Enlaces
  • Tesis en acceso abierto en: TESEO
• Resumen

  • In a world where algorithms are ubiquitous, the development of computational thinking is becoming progressively important among students, technology professionals, and 21st-century citizens in general. Computational thinking has gained importance in the scientific and educational communities over the last decade, and it has been advocated as a fundamental competence that should be included in the compulsory educational curriculum. In addition, several reports have shown that the supply of computer science professionals is not meeting demand in the technological sector. Research has indicated that this problem could be overcome by promoting, from an early age, computational thinking skills that are closely related to computer science and programming. Thus, there is a growing trend to include computational thinking in primary education worldwide due to its many benefits. Educational games as a means of promoting computational thinking have been widely used in recent years. Game-based learning is a type of gameplay with defined learning outcomes that has the potential to provide effective learning experiences for players by including strategies for learning and engagement.. According to research, visual programming, and in particular block-based programming environments, play an important role in introducing K-12 students to the fundamental principles of programming and the computer science world,. For this reason, many block-based games have been developed, and important initiatives to promote computational thinking at a local and international level are based on them. Their broad use increases the need to offer efficient block-based programming environments. This can be achieved by providing block-based programming environments with personalized learning paths through adaptive difficulty.

    The investigation presented herein is focused on offering an adaptive game that offers programming challenges of adapted difficulty based on the learners’ performance. In the course of our research, we proposed an innovative way to define the difficulty of maze-based programming challenges using log data obtained from Kodetu, a block-based maze game. Specifically, we conducted three studies with 9- to 16-year-old learners who were asked to solve sequences of maze-based programming challenges. Using log data from these studies, we investigated the maze characteristics and the coding limitations that affected performance in the challenges and calculated the performance obtained by the participants using a fuzzy rule-based system. The results showed that the turns in a maze, the total number of steps of a maze, and the blocks provided affect student performance. Using regression analysis, we defined a difficulty function for maze-based programming challenges that considers the weights of these factors and provides a first step towards the design of adaptive learning paths for computational thinking-related educational games.

    Having defined the difficulty, we were able to develop the adaptive version of Kodetu, following the approach of the computerized adaptive testing systems. A set of 110 programming challenges was created, comprising the challenge bank of the adaptive Kodetu. These challenges belong to three main categories according to the necessary blocks needed to solve them: sequential, loop, and conditional. The learning path on the adaptive Kodetu is personalized; i.e., depending on how well the learners performed on the previous challenge, the next challenge is provided based on that performance, with the difficulty level dropping or growing. The data obtained from the experiment performed with 9- to 11-year-old learners showed that the adaptive Kodetu is more effective than the non-adaptive version. The comparisons made between the data from the experiments with the adaptive and the non-adaptive Kodetu show that the learners perform better when using the adaptive Kodetu even when the learners playing with the non-adaptive are older. In the sequential and conditional challenge categories, the learners are able to solve difficult programming challenges in less time and with less effort, allowing them to continue playing and developing computational thinking.

    In summary, this investigation presents a novel way to measure the difficulty of block-based maze games and demonstrates the efficiency of learning paths consisting of programming maze-based challenges that adapt the difficulty to the learners’ performance. The proposed difficulty function helps teachers and educational stakeholders to personalize learning paths based on their students’ needs. The use of the difficulty function to develop the adaptive Kodetu and the promising results obtained lead the way to the development and implementation of efficient automated adaptive tools that could be integrated into the curriculum of K-12 education to effectively promote computational thinking.