We are happy to announce the acceptance of a paper to ICSE 2024, co-authored by Alexander Lill, André N. Meyer, and Thomas Fritz.

Software developers dedicate a significant portion of their time to finding answers to coding-related questions. They turn to web searches, Q&A websites, and chat communities for assistance. However, many of their questions have already been answered in chat communities or on Q&A websites. Imagine if there were a way to automatically identify and suggest these existing answers, saving time and effort.

We explored the frequency of recurring questions on the Discord chat platform and developed an automated approach to identify them. We conducted a field experiment, where 277 questions from developers in four Discord communities were answered with links to similar questions, both on Stack Overflow and within the same Discord community. Our results show that:

• 40% of developers found the suggestions helpful.
• Stack Overflow posts were more often considered helpful than previous Discord conversations.
• Developers often prefer describing their issues in natural language questions rather than as search queries.

This research highlights the potential for reusing knowledge from Q&A and chat communities in software development. Differently from current solutions using generative AI, our approach refers to human-written past answers, instead of synthesizing new answers. More details can be found in the full paper: “On the Helpfulness of Answering Developer Questions on Discord with Similar Conversations and Posts from the Past”