Sustainability Flags for the Identification of Sustainability Posts in Q&A Platforms

2025-07-03

AI Summary

Identifying sustainability-related content in software architecture discussions on online Q&A platforms lacks clear guidelines—particularly challenging in cloud architecture contexts. Method: We introduce the novel concept of “sustainability signals”: interpretable, actionable indicators derived from thematic analysis of best practices published by major cloud vendors. Our approach integrates qualitative topic modeling with quantitative controlled experiments, validated on Stack Overflow and other platforms. Contribution/Results: The signal-based framework significantly reduces misclassification (−23.6%), improves classification confidence and accuracy (F1-score +18.4%), and demonstrates strong usability and comprehensibility in user evaluations. Unlike abstract definitions, sustainability signals offer concrete, domain-grounded criteria for identifying sustainability concerns in architectural discourse. This work establishes a new paradigm and reusable methodological framework for mining sustainability knowledge in software architecture.

Technology Category

• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Responsible Web: Sustainability and climate impact of web technologies

Abstract

In recent years, sustainability in software systems has gained significant attention, especially with the rise of cloud computing and the shift towards cloud-based architectures. This shift has intensified the need to identify sustainability in architectural discussions to take informed architectural decisions. One source to see these decisions is in online Q&A forums among practitioners' discussions. However, recognizing sustainability concepts within software practitioners' discussions remains challenging due to the lack of clear and distinct guidelines for this task. To address this issue, we introduce the notion of sustainability flags as pointers in relevant discussions, developed through thematic analysis of multiple sustainability best practices from cloud providers. This study further evaluates the effectiveness of these flags in identifying sustainability within cloud architecture posts, using a controlled experiment. Preliminary results suggest that the use of flags results in classifying fewer posts as sustainability-related compared to a control group, with moderately higher certainty and significantly improved performance. Moreover, sustainability flags are perceived as more useful and understandable than relying solely on definitions for identifying sustainability.

Problem

• Identifying sustainability in software architectural discussions
• Lack of clear guidelines for sustainability recognition
• Evaluating effectiveness of sustainability flags in Q&A platforms

Innovation

• Introducing sustainability flags for Q&A posts
• Flags derived from cloud providers' best practices
• Flags improve sustainability identification certainty