Generation of virtual tutors for PBL-based classes
Virtual tutors, Distance Education, Evasion, k-12 education.
The high rates of failure and avoidance in computing courses become a limiting factor for the development of several professional areas, making the supply of skilled labor scarce. In this sense, courses of game development can mitigate this factor, since they allow to abstract the information and to understand where each concept studied is applied in a playful way. In this context, the use of the PBL model adds value by allowing the student to learn to learn, developing personal characteristics of interpretation and problem solving in an efficient way, as well as skills related to teamwork. However, PBL-based courses have reduced classes to enable the teacher to assist students more effectively. In contrast, this factor limits the opening of more places, due to the cost with teachers, materials per class, among others, both in faceto-face and in distance education. A way to mitigate these costs and improve the efficiency of PBL-based courses through digital virtual tutors. These tutors have features that simulate a teacher and allow them to direct and assist students, however, with more time available, allowing students to learn at their own pace. In this context, note that within a class, students also have different goals and abilities. In this way, the objective of this dissertation is to propose an architecture of generation of virtual tutors for different profiles of students, to be applied in programming courses of games based on the PBL model. In order to evaluate the demand and contributions that this proposal can cause, studies were carried out in order to collect the necessary data to develop the tutors for each student profile and to evaluate the beneficial results that each tutor can bring to each profile. Thus the first study was to develop a systematic review of the literature in order to observe how tutors were developed, mainly related to affective computation. After the results of this study, a prototype was developed for targeted and objective classes based on missions (activities related to the PBL model), and contained in the system an animated virtual character that is currently exhibiting the contents. This prototype was applied to an experiment with middle-level students that allowed the development of a new tutor for a new, more interactive profile of students. This new prototype was developed based on the ALICE chatbot and the AIML language. Upon completion of this tutor, an experiment was conducted using both tutors to evaluate the proposal of a tutors generator to be applied to the same class. This experiment was applied to undergraduate students in order to allow these teachers to replicate the study in their classes in the near future. Finally, the data were analyzed and it was possible to answer the research questions developed during the study. The results presented in this work showed that the students had fun during the classes, motivated by the ludic factor connected to the programming classes of games and virtual tutors. This factor allowed the students to present good results during the development of the games, with more than 75% of the students concluding the problems proposed, with the others presenting around 90% of the problems. Another point observed was that the groups presented a profile distinction in order of 56% to 44% of dispersion among the tutors who most adapted to their profile. These data suggest the possibility of inserting a tutors generator for different profiles of students in this class, however, other studies are needed to show more clearly this presented data. In general, tutors have made classes more dynamic and productive, proving to be excellent tools to support students in the development of PBL-based games.