Towards a Taxonomy of Sustainability Requirements for Software Design

Basic Information

• Paper ID: 2510.08990
• Title: Towards a Taxonomy of Sustainability Requirements for Software Design
• Authors: Mandira Roy (Ca' Foscari University), Novarun Deb (University of Calgary), Nabendu Chaki (University of Calcutta), Agostino Cortesi (Ca' Foscari University)
• Classification: cs.SE (Software Engineering)
• Publication Time/Conference: 2025 ACM Conference
• Paper Link: https://arxiv.org/abs/2510.08990

Abstract

Software systems are significant contributors to global sustainability challenges and require systematic treatment of environmental, social, technical, and economic factors from the initial requirements engineering phase. While existing research provides various sustainability requirements (SRs), these contributions are often fragmented, specific to certain dimensions, or limited to particular application domains, resulting in a critical lack of unified, comprehensive taxonomies within the software engineering community. To address this gap, this study conducts a systematic literature review (SLR) to extract and organize contemporary sustainability requirements. The primary contribution is a comprehensive SR taxonomy spanning four sustainability dimensions (environmental, technical, social, and economic). For each identified category, we provide clear definitions, relevant metrics, and measurements. Additionally, we present a relevance matrix demonstrating positive and negative impacts (synergies and conflicts) across different dimensional categories.

Research Background and Motivation

Problem Identification

1. Core Issue: Software systems' contribution to global sustainability challenges is increasingly significant, yet systematic sustainability requirements taxonomies are lacking
2. Current State Analysis: Existing sustainability requirements research exhibits the following problems:
   • Fragmented research outcomes lacking unified standards
   • Often limited to specific sustainability dimensions (e.g., focusing solely on environmental aspects)
   • Domain-specific without generalizability
   • Lacking systematic organization and classification

Importance Argument

1. Lifecycle Impact: Software's sustainability footprint spans the entire lifecycle, from requirements engineering to deployment and use
2. Design Decision Criticality: Software system sustainability largely depends on its characteristics and design choices
3. Industry Demand: Organizations have recognized this need but lack concrete support and knowledge

Research Motivation

This research adopts a second direction: conducting in-depth analysis of specific software systems to identify and address sustainability issues, aiming to provide the software engineering community with a systematized reference framework.

Core Contributions

1. Systematic Literature Review: Conducted comprehensive SLR to create a sustainability requirements taxonomy
2. Four-Dimensional Taxonomy: Identified comprehensive category sets for four sustainability dimensions (environmental, technical, social, and economic), including definitions, metrics, and measurements
3. Relevance Matrix: Constructed a relevance matrix showing positive or negative impact relationships between different sustainability dimensional categories
4. Open-Source Resources: Provided open-source taxonomy subject to regular updates and extensions

Methodology Details

Task Definition

Input: Sustainability requirements-related literature from various domains and dimensions Output: Structured sustainability requirements taxonomy containing categories, metrics, measurements, and relevance relationships Constraints: Limited to peer-reviewed literature in software engineering and computer science domains

Systematic Literature Review Methodology

1. Research Question Design

• RQ-1: How are sustainability requirements defined and characterized across different domains?
• RQ-2: What are the most commonly identified sustainability requirements categories in existing research?
• RQ-3: What are the commonalities and differences of SRs across different domains?

2. Data Source Selection

• Scopus (www.scopus.com)
• ACM Digital Library (https://dl.acm.org)
• IEEE Xplore (www.ieeexplore.ieee.org)
• Wiley Online Library (https://onlinelibrary.wiley.com)

3. Search Strategy

Employed a three-tier search strategy:

• First Tier: Basic keywords "sustainability" AND "requirements"
• Second Tier: Added dimension-related vocabulary (Social, Economic, etc.)
• Third Tier: Added specific application domains (cloud computing, intelligent systems, healthcare, etc.)

4. Selection Criteria

Inclusion Criteria:

• Top 100 records per query
• Peer-reviewed journal, conference, and workshop papers
• English-language literature
• Computer science or software engineering domains
• Explicitly mention software system sustainability requirements

Exclusion Criteria:

• Research published before 2010
• Broadly discussing sustainability without explicitly addressing requirements
• Non-peer-reviewed research

Taxonomy Construction Methodology

1. Requirements Extraction: Manual analysis of each selected paper to extract SRs and their dimensions and categories
2. AI-Assisted Extraction: Used AI tools to assist knowledge extraction
3. Classification Organization: Organized extracted requirements by dimension and category
4. Deduplication: Ensured taxonomy comprehensiveness and absence of duplication

Experimental Setup

Dataset Construction

• Literature Quantity: Collected through SLR method from four major databases
• Search Result Statistics:
  • Basic search: Scopus 2,863, IEEE 18, ACM 15,822, Wiley 110
  • Social dimension: Scopus 886, IEEE 1,850, ACM 22,285, Wiley 4,735
  • Technical dimension: Scopus 170, IEEE 2,241, ACM 33,088, Wiley 4,735
  • Economic dimension: Scopus 6,321, IEEE 2,897, ACM 14,807, Wiley 0

Relevance Matrix Construction Methodology

1. Dataset Preparation: Constructed dataset containing dimensional categories and their definitions
2. LLM Generation: Used GPT-3.5 and Gemini 2.5 Pro to generate relevance matrices
3. Expert Review: Each author independently marked agreement with generated relevance
4. Result Consolidation: Merged individual author results into single worksheet
5. Group Discussion: Determined final relevance matrix through collective author knowledge and voting

Experimental Results

Primary Taxonomy

Environmental Dimension Categories

• Carbon Footprint: Software carbon intensity metrics
• Energy Consumption: Energy efficiency, runtime efficiency, CPU intensity, memory usage, etc.
• Electronic Waste: Number of non-reusable modules
• Code Sustainability: CPU usage, memory usage, code smells
• Resource Management: Energy-efficient data management, water footprint

Technical Dimension Categories

• Persistence: Technology evolution, functional evolution
• Longevity: Mean time between failures, average service life
• Repairability: Cohesion and coupling, documentation traceability degree
• Feature Bloat: Percentage of user engagement with various features
• Security Issues: Number of known vulnerabilities

Social Dimension Categories

• Digital Inclusion: Number of supported languages, demographics
• Fair Access: End-user data availability, horizontal and vertical equity
• Ethical Issues: Degree of bias
• Customer Social Value: Number of positive feedback items
• Social Engineering: Well-being
• User Error Protection: Recovery time

Economic Dimension Categories

• Circular Economy: Code or service reuse percentage, number of shared infrastructure
• Affordability: Access cost
• Cost Efficiency: Development cost (number of workers, time, and effort)
• Software Process Evolution Knowledge Capital: Customer capital value, market demand value, etc.

Relevance Analysis Results

Positive Correlations (Selected Examples)

• Carbon Footprint↓ ⟷ Energy Consumption↓: Lower energy usage reduces greenhouse gas emissions
• Energy Consumption↓ ⟷ Code Sustainability↑: Optimized code reduces computational power and energy requirements
• Electronic Waste↓ ⟷ Circular Economy↑: Reuse and recycling reduce waste volume and toxicity
• Digital Inclusion↑ ⟷ Fair Access↑: Removing access barriers helps more people use digital systems

Negative Correlations (Selected Examples)

• Feature Bloat↑ ⟷ Energy Consumption↑: More features increase resource usage and energy demands
• Feature Bloat↑ ⟷ Longevity↓: Overloaded systems become obsolete faster
• Security Issues↑ ⟷ Affordability↓: Stronger security may increase costs
• Digital Inclusion↑ ⟷ Cost Efficiency↓: Inclusion may increase development costs

Experimental Findings

1. Dimensional Balance: All four sustainability dimensions contain rich categories and metrics
2. Metric Diversity: Metric types include numerical, percentage-based, and qualitative
3. Relevance Complexity: Complex positive and negative correlations exist across dimensions
4. Context Dependency: Relevance relationships may vary based on specific contexts or scenarios

Related Work

Primary Research Directions

1. Frameworks and Methodologies: Such as ShapeRE theoretical framework, goal-based scenario modeling approaches
2. Dimension-Specific Research: Research focusing on single sustainability dimensions (environmental, social, or economic)
3. Domain-Specific Applications: Sustainability requirements in specific domains such as e-learning systems, healthcare applications, intelligent systems
4. Empirical Research: Such as studies on linguistic relativity theory's impact on SR identification

Advantages of This Work

1. Comprehensiveness: Covers four sustainability dimensions rather than single dimensions
2. Systematicity: Systematically organizes existing knowledge through SLR methodology
3. Practicality: Provides concrete metrics and measurements rather than theoretical frameworks only
4. Openness: Open-source taxonomy supporting continuous updates and extensions

Conclusions and Discussion

Main Conclusions

1. Taxonomy Establishment: Successfully constructed comprehensive taxonomy spanning four sustainability dimensions
2. Relevance Revelation: Revealed complex relationships between different sustainability categories through relevance matrix
3. Practical Guidance: Provided software developers with concrete sustainability requirements reference and trade-off guidance

Limitations

1. Temporal Constraints: Survey work limited by time, potentially missing some relevant research
2. Domain Limitations: Search restricted to specific domains, potentially incomplete coverage
3. Subjectivity: Relevance matrix determination based on expert judgment with inherent subjectivity
4. Context Dependency: Relevance relationships may vary based on specific product contexts

Future Directions

1. Taxonomy Extension: Continuous updates and extensions of open-source taxonomy
2. Empirical Validation: Validate taxonomy effectiveness in actual software projects
3. Tool Development: Develop automated tools supporting sustainability requirements engineering
4. Cross-Domain Research: Extend to more application domains and emerging technology areas

In-Depth Evaluation

Strengths

1. Rigorous Methodology: Employs standard systematic literature review methods ensuring scientific rigor and reproducibility
2. Practical Contribution: Constructed taxonomy has direct practical application value, filling important domain gaps
3. Comprehensive Perspective: Simultaneously considers four sustainability dimensions, avoiding single-dimension research limitations
4. Strong Innovation: First systematic construction of cross-dimensional sustainability requirements taxonomy and relevance analysis
5. Open Sharing: Open-source publication of research results promotes academic community knowledge accumulation

Weaknesses

1. Insufficient Validation: Lacks application validation in actual projects; connection between theory and practice needs strengthening
2. Limited Quantitative Analysis: Relevance analysis primarily based on qualitative judgment, lacking quantitative statistical validation
3. Insufficient Dynamic Consideration: Inadequately addresses dynamic changes in sustainability requirements over time and technological development
4. Missing Tool Support: Lacks concrete tools or methods for applying the constructed taxonomy

Impact

1. Academic Contribution: Provides important theoretical foundation and reference framework for sustainable software engineering field
2. Practical Value: Provides software development organizations with systematic sustainability requirements guidance
3. Standardization Advancement: Facilitates standardization of sustainability requirements engineering processes
4. Interdisciplinary Promotion: Promotes interdisciplinary fusion of software engineering and sustainability research

Applicable Scenarios

1. Requirements Engineering Phase: During software project requirements analysis and specification definition
2. System Design Phase: In architecture and detailed design considering sustainability factors
3. Project Assessment: For evaluating existing systems' sustainability levels
4. Research Reference: As foundational reference material for sustainable software engineering research
5. Education and Training: For teaching sustainability concepts and practices in software engineering education

References

The paper cites 23 important references, primarily including:

• Karlskrona Manifesto for green software engineering 5
• Key research on sustainability requirements engineering 4, 6, 7, 9
• Domain-specific sustainability requirements research 2, 10, 16, 17
• Systematic literature review methodology 13
• Large language model technology 12, 23

Overall Assessment: This is a high-quality research paper with significant contributions to the sustainable software engineering field. Through systematic literature review methodology, the authors successfully constructed a comprehensive sustainability requirements taxonomy, filling an important gap in the field. While there remains room for improvement in empirical validation and tool support, its theoretical contributions and practical value are substantial, providing important foundational support for advancing sustainable software engineering development.