Investigando o potencial do uso de exemplos trabalhados no auxílio a solução de issues em um projeto no GitHub

Abstract

The recent increase in contributions to open-source software projects, especially on GitHub repositories, raises questions about the difficulties faced by beginner programmers during the contribution process, which often leads to insecurity due to unfamiliarity with the project, its community, and adopted development standards. Even with attempts to alleviate these challenges for inexperienced developers through mentorship, good community and documentation practices, several specific hurdles remain. In this context, the use of Worked Examples emerges as a potential alternative to assist developers during the contribution process in open-source projects. The objective of this study is to investigate how the use of Worked Examples of different similarities, presented as closed and approved issues and automatically recommended by a bot based on an initial GitHub issue, can influence the resolution of issues from JabRef by Bachelor of Computer Science students. The study also aims to identify the benefits and challenges resulting from adopting this practice. Concurrently, by monitoring participants’ eye gaze with an eye-tracker, the research seeks to evaluate the cognitive effort expended during issue resolution using established literature metrics and to identify relevant criteria for selecting more similar Worked Examples. The results indicate that high-similarity Worked Examples, especially when presented in groups of three recommendations, helped study participants formulate solutions for the proposed issues, serving as a theoretical basis for understanding the problem. This is evidenced by the analysis of eye-tracker data, which showed a significant number of Areas of Interest (AOIs) in the recommendations that captured participants’ attention and were useful in some way for formulating a solution. However, this may also indicate a high dispersion of attention during the analysis of recommendations. Furthermore, the complexity of recommendations combined with the similarity between the Worked Example and the proposed issue remain important challenges to overcome. The findings of this study provide guidelines to enhance Worked Example recommendations, aiding in developer training and inspiring new approaches in open-source software communities.