The research was conducted at the Sofia High School of Mathematics, in specialized informatics classes, involving 10th and 11th grade students. The study focused on identifying the benefits and effects of using automated grading systems in programming education, particularly regarding students’ confidence, problem solving, and metacognition. It examined whether the identified patterns support the trend of moving towards more personalized and adaptive informatics learning.

The study involved more than 70 students, who used the HackerRank platform to complete over 30 practical Java programming tasks. Four tasks were examined in more detail, based on two criteria: the percentage of students who solved the task correctly, and the percentage who did so on their first try. These metrics allowed the comparison of students’ actual and perceived results.

To increase the validity of the quantitative data, qualitative data was gathered through interviews and questionnaires. The students were asked to share their perception of the automated system, whether they believed that their answers were correct when submitting them, and how the system’s feedback compared with their perception. This triangulation revealed discrepancies between the two and emphasized the role of feedback in learning.

The results also revealed that many students were quite confident until they met the failed test cases. This inconsistency implies that automated feedback may help to decrease overconfidence and increase metacognitive development. Some of the common mistakes included algorithmic errors and Java-specific syntax errors. Therefore, the study proposes a way of constructing an effective test case set based on typical error patterns to ensure that both comprehension and application are assessed.

Furthermore, the study analyzed students’ behavior after incorrect submissions. Some tried again, others looked for alternative ways to solve the problem. These behaviors show students’ persistence and cognitive flexibility when feedback is instant.

The automated grading system allows teachers to pay more attention to those students who need individual assistance, while students are able to learn at their own pace. The analysis points to the possibility of developing an adaptive algorithm that will be able to choose future tasks of a similar level of difficulty, which will make it possible to create individual learning plans. Teachers act as mentors in a model that shifts from conventional, uniform class assignments to individualized learning.

Even though the research was carried out in the context of programming, the results are applicable to other STEM subjects that require algorithmic thinking and feedback. It is possible to implement this strategy regardless of the language or curriculum, as it provides scalable personalized feedback and reduces the time spent on grading. In conclusion, automated grading tools can improve informatics education to become more student-centered.

The article is structured as follows:
1) Introduction,
2) Methodology,
3) Data collection,
4) Data analysis and Interpretation, and
5) Conclusions.

Keywords: Automated grading, programming education, informatics teaching, student feedback, Java, error analysis, adaptive learning, secondary education, STEM, personalized instruction, teachers’ training.