What is it about?
Good software depends on good communication and collaboration. When these go wrong, teams develop community smells—recurring negative patterns in how people interact—which add up over time as social debt. Machine learning (ML) projects are a good place to study them, because people with different roles, such as data engineers and software engineers, must work together. In this study, we identified which community smells appear in ML-enabled systems, what causes them, what effects they have, and how to mitigate them. We built hypotheses from the literature and from interviews, and tested them with a questionnaire and a statistical technique (PLS-SEM).
Featured Image
Photo by Vitaly Gariev on Unsplash
Why is it important?
We built and validated five models, one for each of five community smells, that link causes, effects, and mitigation strategies. Practitioners can use these models to recognize and address community smells in their own organizations, and researchers can use them to study the social side of ML development.
Perspectives
Machine learning projects bring together people with different backgrounds, such as data engineers and software engineers. This makes them a good setting to study community smells—recurring problems in how people communicate and collaborate. With this paper, we tried to go beyond describing the smells: for five of them, we modeled their causes, their effects, and the strategies that can reduce them. I hope practitioners can use these models to look at their own teams.
Prof. Stefano Lambiase
Aalborg Universitet
Read the Original
This page is a summary of: Uncovering Community Smells in Machine Learning-Enabled Systems: Causes, Effects, and Mitigation Strategies, ACM Transactions on Software Engineering and Methodology, January 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3712198.
You can read the full text:
Contributors
The following have contributed to this page
