Fang, Zihan; Huang, Yu. “‘Math is a pain!’: Understanding challenges and needs of the Machine Learning community on Stack Overflow“. Proceedings of the ACM on Human-Computer Interaction, vol. 8, no. CSCW2, 2024, p. 451. DOI: 10.1145/3686990.
Stack Overflow (SO) is a popular online platform where people can ask and answer programming questions. It has also developed a community focused on machine learning (ML), where both beginners and experts can share ideas and solve problems. However, research shows that engagement in the ML community on SO is lower compared to other topics, and it’s not fully understood why this happens. This study uses 22 hours of interviews and 131 survey responses from SO users to identify the main reasons behind the low response rate and long wait times for ML questions. Key factors include the high level of difficulty in posting good ML questions, the difference between the time spent asking and answering questions versus the benefits gained, the scattered nature of the ML community across different platforms, and suggested improvements for SO. The study also finds that SO’s culture has become less welcoming over time, with newcomers experiencing stress when posting or answering ML questions—though this stress lessens as they gain more experience. Finally, the study looks at how generative AI tools like ChatGPT might affect ML Q&A on platforms like SO. The findings aim to improve the experience for ML users on online platforms and encourage better knowledge-sharing and collaboration in the ML field.
