Enseñanza, evaluación y análisis de habilidades de pensamiento computacional en etapas tempranas

Computational Thinking (CT) is a cognitive skill that must be acquired from an early age in order to adapt to today's computerised society. Many of the abilities commonly associated with CT can be developed by learning to program, such as abstraction, problem decomposition, algorithms, and program debugging. The learning and assessment of skills that contribute to CT development in formal education is of increasing importance on an international scale. Moreover, there is a consensus that this learning should start at an early age. In order to include the teaching of programming and the development of CT skills in the school curricula, it is necessary to determine what should be learned at what age and how. Being such a recent competence, the definition of the concept is still unclear, as well as the operational frameworks for effective teaching and assessment. Furthermore, there are no validated and agreed instruments for the learning and assessment of CT. In addition, it is necessary to establish which skills are achievable and whether there are differences in the acquisition of these skills according to the age and gender of the children.

In this thesis, a systematic review of the literature on theoretical and operational CT is carried out in order to provide a suggested definition and decomposition of CT, as well as hypothesis about systems that combine different strategies for the learning and assessment of CT. From the state of the art, it can be found that new assessment tools have emerged in recent years, and these are not usually oriented to early stages. At the time of the present study, there was no stand-alone assessment test (not related to a specific learning environment) for primary school students in the scientific community.

For this reason, we designed and validated, through an experience in schools, a standalone test that would provide a correct assessment of CT in early stages. This test: Beginners Computational Thinking test (BCTt) was validated through a case study with primary school students, with very successful results in terms of reliability and with wide international acceptance, establishing several connections and starting collaboration projects for the improvement and development of new instruments for the assessment of CT. These collaborations are oriented towards a common interest in obtaining assessment instruments that cover the educational range from pre-school to higher education. In addition, the BCTt has been translated into multiple languages, and several experiences have been carried out in different populations and countries (such as Portugal, Belgium or Singapore). Future lines of action have also been proposed and are in different stages of development.

Since the first questions of the BCTt were too easy for the last stage of Primary Education, a new adaptation of the test for these grades, the competence Computational Thinking test (cCTt), was developed in collaboration with the Ecole Polytechnique Fédérale de Lausanne (EPFL) in Switzerland, through a case study in several schools in Switzerland. Morevoer, a case study was also carried out, in collaboration with the EPFL in Switzerland and the NGO Tree Tree 2 in Portugal, to establish the age limits for using the BCTt and the cCTt, in 3rd and 4th grades of Primary Education, where there is doubt as to the application of one or the other test.

In order to establish the lower age limit for the administration of the BCTt, a new case study was carried out in Early Childhood Education, based on the teaching of programming through robotics, in collaboration with Fontys University of Applied Science and Zuyd University of Applied Science, both in the Netherlands. This experience made it possible to validate the BCTt for children aged 4 and 5. In this way, the BCTt and the cCTt are validated as autonomous instruments, independent of any environment, and reliable for the assessment of CT in Primary Education and, therefore, can be administered as a pre-test and post-test in the investigations that require it.

Within the context of the thesis, and in order to explore new CT learning and assessment strategies, the collaborative video game Blue Ant Code (BAC) has been developed, as well as several modules for game data analysis and filtering. This tool is based on game-based learning for the learning and assessment of CT through learning analytics. A case study was conducted in three primary schools in which BAC was validated using the previously developed BCTt. Through this study, a system that combines different CT learning and assessment strategies, was also validated. The results show that this system is suitable, especially at early ages, covering the three key dimensions of CT (computational concepts, practices, and perspectives). The results obtained were very strong, resulting in a significant improvement in learning and retention over time. BAC also seemed to be better suited to younger students and those with special needs or low percentile. One of the most relevant findings is that students in the first years of Primary Education were able to outperform students in higher grades in CT skills at the end of the study, which highlights the importance of developing CT from a very early age, and suggests the incorporation of certain computational concepts, such as nested loops or conditionals, into the school curricula in the first year of Primary Education or even earlier.

In order to be able to assess, in addition to strictly academic skills, also aspects related to intrinsic motivation, interests, persistence and behaviours in response to rewards for learning programming and CT, a version of BAC, with new functionalities, was developed for large-scale deployment in a non-school and voluntary environment. In this massive study, more than 28,000 games were examined to analyse the behaviour of users who voluntarily played with BAC. Through game-based learning analytics, relevant findings were also drawn such as that, in a non-school context of voluntary play, the computational concepts of simple, nested, and conditional loops can be mastered between the ages of 3 and 6, which could be an aspect to take into account for the development of school curricula. Moreover, it is noteworthy that all these concepts can be fully mastered by the age of 4 if the children persist in a challenge on a voluntary, intrinsically motivated basis.

On the other hand, other findings can help to guide the design of improved learning tools for programming and CT development. Since age and gender have an impact on performance, it is advisable to adapt learning tools accordingly, as well as the school curricula. In addition, learning environments could adapt the way reinforcement and rewards are provided, especially for boys for more complex challenges and for girls from the age of 5, as significant differences in these behaviours were identified according to children's age and gender.

In overall, all the work carried out and presented in the thesis has enriched the state of the art in the new field of Computational Thinking and especially regarding teaching-assessment tools and the specification of age/gender ranges and behaviours in the learning of this competence.