How AI Developers Discuss Sustainability Concerns in Practice
Explore how AI developers talk about energy, carbon, and sustainability trade-offs, and what this means for better tools and practices.
Sustainability in AI is not only a technical issue. It is also a communication issue: what developers discuss, what they ignore, and how teams justify trade-offs under real delivery pressure.
The Problem
AI developers make many decisions that influence energy use and carbon impact: model choice, infrastructure setup, prompt strategy, retraining frequency, and deployment configuration. Yet we still know relatively little about how developers actually discuss these sustainability concerns in day-to-day work.
Without that understanding, it is difficult to design effective guidelines, metrics, and tools that fit real development practices.
What This Thesis Is About
This thesis explores how AI developers discuss sustainability concerns in practice across technical and social spaces (for example issue trackers, pull-request discussions, forums, and documentation). The goal is to identify recurring themes, tensions, and decision patterns around energy and carbon trade-offs in AI development.
This is a strong fit for students who enjoy combining AI/software engineering with empirical research that has direct impact on real-world practice.
Why This Topic Is Exciting
- You work at the intersection of AI, sustainability, and developer behavior.
- You gain mixed-method research skills (qualitative coding plus quantitative analysis).
- You produce insights that can directly inform greener engineering workflows and tool design.
Objectives
- Map sustainability discourse patterns in AI developer communities and projects.
- Identify barriers and enablers that shape whether sustainability concerns are raised and acted upon.
- Analyze how trade-offs are framed between performance, cost, speed, and environmental impact.
- Propose actionable recommendations for teams, maintainers, and tool builders to better support sustainability-aware AI development.
What You Will Do
| Phase | Work |
|---|---|
| Scoping & protocol | Define research questions, data sources, and coding framework |
| Data collection | Collect and curate developer discussions from selected AI projects/communities |
| Analysis | Apply qualitative/quantitative analysis to identify themes, patterns, and trade-offs |
| Validation | Triangulate findings via comparison across projects or small follow-up interviews/surveys |
| Outputs | Deliver thesis, curated dataset/coding schema, and practical recommendations |
What You Gain
- Experience with high-value empirical methods used in software engineering and HCI-style research.
- A strong portfolio piece combining AI, sustainability, and developer studies.
- Skills in translating evidence into practical guidance for research and industry audiences.
Who Should Apply
Required courses:
- Programming in Python (INF-22306)
- Machine Learning (FTE-35306)
- Big Data (INF-33806)
Required skills & mindset:
- Solid analytical skills and clear writing
- Interest in AI systems, sustainability, and developer practices
- Curiosity for empirical research (qualitative and quantitative)
- Motivation to work independently while discussing progress regularly
Key References
- Verdecchia, R., Sallou, J., & Cruz, L. (2023). A systematic review of Green AI. WIREs Data Mining and Knowledge Discovery, 13(4), e1507. doi:10.1002/widm.1507
- Henderson, P., Hu, J., Romoff, J., Brunskill, E., Jurafsky, D., & Pineau, J. (2020). Towards the systematic reporting of the energy and carbon footprints of machine learning. Journal of Machine Learning Research, 21(248), 1-43.
- Strubell, E., Ganesh, A., & McCallum, A. (2019). Energy and policy considerations for deep learning in NLP. ACL 2019, 3645-3650.
- Luccioni, A. S., Jernite, Y., & Strubell, E. (2024). Power Hungry Processing: Watts Driving the Cost of AI Deployment? FAccT ‘24. doi:10.1145/3630106.3658542