Computer science is an important subject not only for those who want to become computer scientists but also for everyone because it develops necessary skills for different fields. Upper secondary school is a critical time for making the choice regarding further studies. However, students' choices can be influenced by the organization of computer science (CS) education in upper-secondary schools or the availability of opportunities to study CS. As schools in Estonia have a high degree of autonomy, the aim of this research is to give an overview of the organization of CS education at the upper secondary level in Estonian schools and find out how it influences further studies in the information technology (IT) field. Three research questions were set for the research: (1) How is the teaching of computer science organized in different upper secondary schools, and what kind of additional activities are conducted to support it? (2) What support elements and obstacles are identified by school leaders in computer science education? (3) Are the supporting and hindering factors in school related to further study in the field of IT? The sample of the research consists of 30 different upper secondary schools (18.5 % of all schools of this type) in Estonia. Data was collected from the schools' curricula and school principals' questionnaires. The SPSS program was used for data processing and chi-square test and correlation analyses were performed. The study revealed that the organization of computer science education can be very variable across schools in Estonia. There are schools with compulsory courses, schools with elective courses, and those with both options. There are schools where computer science is not taught at all and schools where 7 different CS courses are offered. More than half of the schools cooperate with universities, and half of the schools get support to teach CS from the local government. A lack of qualified teachers and quality learning content was identified as hindering factors by both, the schools that teach computer science and those that do not. Principals of the schools where CS is not taught are significantly more concerned about the curriculum being overloaded, as opposed to those of the schools where CS is taught. On average, 13 % of all students choose IT as their further field of study. It was found that the more computer science courses a school offers, the more students proceed to study IT. In addition, students from the schools that cooperate with universities and other partners are more likely to continue their studies in the IT field. On the other hand, significantly fewer students proceed with their studies in IT if they come from schools where principals identified obstacles such as insufficient technical bases, missing technological support, overloaded curriculum, and limited interest of students. The results can be useful in understanding how different approaches to teaching computer science can affect the choice of further studies and what can be done, based on this, to improve computer science education in upper secondary schools.

Due to the pandemic, learning transformed from traditional face-to-face modes through emergency remote to distance and blended or hybrid learning. The choice of the form of teaching-learning determines how exactly studies are carried out and what role the teacher has. It is crucial to identify the optimal choice among these learning options as each offers its own set of advantages, and both students and teachers have their individual preferences. The aim of this paper is to ascertain the advantages and disadvantages of different forms of learning and examine the variations in learning outcomes in the context of different learning forms. A study was conducted by analyzing academic performance over five academic years, from 2018 to 2022, in the course “Object-oriented programming”, where the grading system encompasses weekly programming assignments, quizzes based on lecture material, a group project, two programming tests and a final examination. Based on the points obtained in the course, the highest results were achieved in 2020 when emergency remote teaching was used. The results for 2019, which was the last ‘normally’ conducted year, were the weakest, while those for 2021 with distance learning were back to their pre-COVID levels. Offering a possibility to choose between in-class and distance education led to improved results in 2022. The comparison of the various learning forms revealed that there is no one optimal method. However, it can be speculated that blended and hybrid learning will displace face-to-face learning.

Success in MOOCs can be defined differently, either in terms of course completion or in terms of fulfilling learners' intentions. This paper proposes a model of using both approaches to evaluate learners’ success and identifies factors influencing them. The scoping literature review was conducted and the expectancy-value theory was integrated to create a comprehensive model. Several factors, e.g. gender, education level, learner's previous experience and earning certificates, affect both kinds of success. While factors affecting completion-based success have been quite well examined, the scoping review underscores the need for future studies to deepen the understanding of the less-explored aspects of learners' intentions-based success.

In academia, plagiarism is a critical concern, and educators require effective tools to identify and prevent it. Presently, many plagiarism detection tools rely on source code comparison, which may not effectively counter the obfuscation techniques employed by students. This article presents an innovative solution to identifying potential plagiarism in programming assignments through the analysis of logs containing information on user actions during the programming process. The created tool adopts a history-based approach to plagiarism detection, which helps to counteract certain forms of obfuscation students use to conceal their plagiarism. The plagiarism detection tool analyses logs based on user-specific criteria such as run count, total time spent working, log file size, and pasted text ratio. The tool also compares log files for detection of duplicate files, identical texts pasted in different log files, source code pasted in different log files, and source code similarity. The solution also allows the user to specify the values for each analyzed metric for plagiarism detection. The effectiveness of the tool is demonstrated through experimental evaluations, enabling to identify cases of plagiarism that could not be detected with other available tools. The findings suggest that the tool can be an efficient and effective means for educators to identify plagiarism in programming assignments.

This article introduces and reflects on a new user-friendly system for creating and maintaining automated assessments built on an existing platform, called Lahendus. The central component of the new system is the user interface that is the bridge between teaching staff and automated assessment code. Noteworthily, the user interface relies on Test Specific Language for defining test cases and on a dedicated testing library, called Tiivad. These components allow for both dynamic and static assessments. The system has been used with success in multiple introductory programming courses. The article reflects on the efficiency gains and simplifications made by comparing the new system to the previous one. The biggest gains came from the reduction of complexity and time required for creating automated assessments. The system has been used only a little over three months, but it already showcases substantial time and effort savings. During that period, automated assessments for more than 100 programming tasks were created, and more than 170 hours were saved in the process of creating automated assessments. The results showed that the new system alleviates teaching staff's high workload.

The computer programming MOOC “From Technology Consumer to Creator”, which is under consideration in this article, was created in 2020 with the aim of expanding the possibility of learning programming in Estonian schools. It is intended for young people who have no prior programming experience. The project-based learning phase is included at the end of the course to recall the learned knowledge and give an opportunity to gain new skills during this phase. The aim of this paper is to provide an overview of the strengths and weaknesses of the projects carried out during the MOOC held in the spring of 2020. 469 students participated in this MOOC, and their 197 project solutions were reviewed. The results show that the students preferred to do the projects individually, however, their programs had on average the same number of rows as the programs done in teams. Although the code written was not completely without errors and had some redundancies and repetitions, the students acquired new knowledge and gained new experience from doing the project.

The course “Introduction to Databases” focuses on teaching Structured Query Language to non-IT students. As the learners have had no exposure to programming-related courses or programming languages and have various backgrounds, it can cause different problems, and the students might need help while solving homework assignments. The goal of this paper is to give an overview of creating troubleshooters for the introductory database course. A troubleshooter is a set of hints designed to give clues to help the student overcome the most common problematic parts of the task. Interviews with course instructors, forum posts and students' individual homework assignments were analysed for creating six troubleshooters. Statistics from the troubleshooter environment and ratings by learners were collected to assess the quality of the development work. Although the learners did not rate the created troubleshooters highly, they agreed that the use of the troubleshooters should be continued in the following years. The comments of students help to improve the troubleshooters for the next courses.

Investigation of the programming process can provide teachers with valuable information regarding the practicalities of teaching and diverse programming styles. When grouping learners according to behavioral patterns, it is essential to determine the characteristics of the learners with the lower exam scores. Using this knowledge, the teaching and learning processes can be made more effective. Considering the above, this article poses two research questions: 1) What types of solvers can be differentiated from the analysis of the programming process? 2) Are there statistically significant differences in the midterm exam scores of different solver types? The data for the analysis were collected from the log files of the students' first programming course. Four features were calculated and used in the analysis: the number of program executions, the number of error messages, the percentage of typed characters of the program at the first run, and the percentage of syntax errors. Four types of solvers were identified using the k-means cluster analysis: 1) Frequent pressers of the run button; 2) Receivers of syntax errors; 3) Balanced solvers; 4) Late starters of the program execution. Using Kruskal-Wallis and Mann-Whitney U tests, it was observed that the “Receivers of syntax errors” and the “Late starters of the program execution” had statistically significantly lower scores on the midterm exam than the “Balanced solvers”. The study reveals the importance of program execution at an early stage of the programming process. It could indicate that teaching needs to focus more on testing and debugging programs, which helps students understand programs better and find mistakes faster. The learners who struggle through the programming process should be guided to write programs incrementally, as this can also help find and correct syntax errors more quickly.

When the flipped classroom approach is used in a course with a large number of students with different backgrounds, some students can struggle with learning and completing homework before the class. This paper presents the so-called “troubleshooters”, which can provide students with additional support in solving homework exercises. Troubleshooters are help systems, structured like decision trees, that contain hints and answers to questions that can emerge when solving a particular exercise. The goal of this paper is to give an overview of troubleshooters and to investigate students’ opinions about them. The results show that students have used troubleshooters and have found them helpful. Also, students indicated that troubleshooters helped them understand more difficult parts of the exercises and the availability of troubleshooters reduced, for some students, the need to ask the instructor for help. Furthermore, a relationship between the use of troubleshooters and the final results in the course was explored. The results indicate that students with lower scores used troubleshooters more and reported receiving more support from troubleshooters. The development of troubleshooters for course exercises has been beneficial both for the students and instructors as they can help students persist in the course, lower the instructors’ workload, and provide more time for other activities.

Information technology (IT) has been a popular specialty among high-school graduates for quite some time. As the number of entrants continues to grow year-by-year, the load on teaching staff also increases. Automation of teaching activities can be used to alleviate the high workload resulting from courses with hundreds of students. The aim of this study is to ascertain the processes that could be automated in the context of assessment and feedback. To gather ideas, eight one-hour mini-group interviews were conducted at the Institute of Computer Science, the University of Tartu, with 17 faculty members responsible for teaching IT courses. A qualitative approach was used for classifying the proposed ideas related to the automated assessment and automated feedback. Five main categories connected to automated assessment emerged from analyzing the codes generated based on the interviews using qualitative content analysis: use of automated assessment, instructor-related topics, negative effect on students, improvements of automated assessment, and technical concerns. Regarding automated feedback, three main categories emerged: feedback for teachers, feedback for students, and feedback for both teachers and students. The classification of ideas was used as an input for designing a schema displaying the current automated assessment system with all the potential additions suggested by the faculty members showing how, if and where exactly the proposed ideas would fit. We believe that the ideas and improvements related to automated assessment and feedback in this study, in conjunction with the detailed clarification of the process, can be a useful input for other institutes. The results will be used as a basis for implementing and improving the automation tools used hand in hand with teaching by the faculty members.

Solving individual tasks plays an important role while studying programming but teachers do not see how learners achieve the result, especially in the case of massive open online courses (MOOCs). Investigating the process of solving programming tasks can provide the teacher with valuable information. The aim of this study is to find answers to the following questions based on the log files of the programming environment Thonny: What behaviour patterns of different learners in solving programming tasks can be identified based on the example of a summary task of the MOOC “Introduction to Programming”? What types of learners can be distinguished on the basis of the behaviour patterns of task solving? The analysis of learners’ behaviour patterns was performed by observing whether they solved the task gradually or not, i.e., whether they wrote the program in parts while also testing their work in the meantime, or whether they tested it for the first time only after they had written the bulk of the solution. In the context of the solution of the MOOC summary task, one can observe that the seven learners who spent the most time and had the greatest number of error messages and repetitive error messages did not work in a gradual style. It is possible that recommending a gradual solution process can help beginners to better understand their code and find errors. It can reduce situations with many concurrent errors in the code making it difficult to understand why any specific error message was generated.

A large number of students drop out in the first year of their Bachelor's studies in Computer Science (CS). It is a problem in many countries worldwide. The aim of this study was to develop a scale for measuring the motivation and perception factors that influence students' decision to start studying Computer Science (CS), and to find the differences between these factors. A questionnaire was conducted based on expectancy-value theory to measure students' motivation and perception. Data were collected from 140 students who started their Bachelor studies in CS at a university in Estonia. The scale, named FICSS (Factors Influencing Computer Science Studies), was conducted and validated using exploratory and confirmatory factor analyses. Two models were identified: a four-factor model describing students' motivation ("Intrinsic value", "Utility value", "Social influences" and "Time for family") and a five-factor model ("Satisfaction with choice of specialty", "Social dissuasion", "Demanding work", "Salary and job security" and "Social status") that describes students' perception. "Intrinsic value" and "Utility value" were the highest-rated motivational factors. The lowest-rated factor was "Social influences". Among the perception factors, the factors with the highest means were "Satisfaction with choice of specialty" and "Salary and job security" while "Social dissuasion" was rated lowest.

T-algebra is an interactive learning environment for step-by-step solving of algebra problems. To make the diagnosis of mistakes more complete, each solution step in T-algebra consists of three stages: selection of the transformation rule, marking the parts of expression, entering the result of the operation. This article describes the last stage that can proceed in three different modes: free input, pattern-driven input and input of some components of the result.