An Empirical Study of Unanswered Python-Related Questions on Stack Overflow

As a high-level programming language, Python supports user-friendly coding and system integration. It is not only used for data analytics but is also applied in software development. Although it has several benefits, users have various issues while running Python code. When programmers face issues, they often use Stack Overflow (SO) to get assistance for their problems. However, many Pythonrelated questions on SO are not responded to and remain unanswered. Thus, in this paper, we analyzed these unanswered questions to identify potential weaknesses of user and discussion characteristics. We applied a mixed-methods approach, including quantitative and qualitative analysis of user reputation, question types, and discussion topics. Our analysis shows that unanswered questions are mostly from low-reputation users ($\mathbf{7 2. 5 2 \%}$), followed by mid-reputation users $(\mathbf{1 7. 9 3 \%}$), and high-reputation users $\mathbf{(9. 5 5 \%)}$. The distribution of the unanswered question types shows that low-reputation users ask more “how” type of question ($\mathbf{3 8. 9 0 \%}$) followed by mid and high-reputation users. In contrast, the “what” and “why” questions are mostly posted by high-reputation users, followed by mid and low-reputation users. Our word-cloud analysis suggests a correlation between user reputation and the question topics. We acknowledge that there are limitations in our study, include potential misclassifications due to the time of posting and biases in manual labeling, which may affect the accuracy of our findings. This study provides insights into unanswered questions posted by the community, open up potential areas for future work such as improving support mechanisms and community engagement.