Hardware choices determine who pays the energy bill, who gets access, and whose data is at risk — yet ethical questions rarely enter the systems design conversation. How do we build infrastructure that is both technically excellent and socially responsible?
The Workshop on Ethical Systems and Architecture Design (HotEthics'26) brings together researchers working across the full system stack — from microarchitecture to datacenter infrastructure — to identify where design choices create or exacerbate ethical problems, and to develop system-level solutions. Building on the inaugural 2024 workshop, HotEthics'26 is a forum for sharing early-stage work, challenging prevailing assumptions, and forging connections across systems, architecture, and the broader societal-impact community.
Program
Afternoon Block 1 13:30 – 15:30
| Time | Session / Talk | Authors / Affiliations |
|---|---|---|
| 13:30 – 13:45 | Opening Remarks | — |
| Session 1: Economic & AI Realities (45 min) | ||
| 13:45 | Big Tech: An Industry with an Increasingly High Energy-Revenue Intensity | (Ghent University, Belgium) |
| 14:00 | Tracing the Computer System Supply Chain | |
| 14:15 | Characterizing the Intergenerational Carbon Footprint of AI Infrastructure | (Ghent University, Belgium) |
| Session 2: Sustainable Hardware & Efficient Infrastructure (60 min) | ||
| 14:30 | Do Energy-Harvesting Pixels Reduce Carbon Cost of Visual Computing Systems? | (University of Rochester); (Northeastern University); (University of Rochester); (Northeastern University); (University of Rochester) |
| 14:45 | Bio-ReRAM Intelligence at the Edge Can Help Energy Efficiency and Sustainability for the Environment | (University of Illinois Urbana-Champaign) |
| 15:00 | The Need for Computational Pluralism | |
| 15:15 | Fleet-Level Optimization for Heterogeneous Device Provisioning in Machine Learning Model Deployments | (Carnegie Mellon University) |
15:30 – 16:00 — Coffee Break
Afternoon Block 2 16:00 – 18:00
| Time | Session / Talk | Authors / Affiliations |
|---|---|---|
| Session 3: Reliability, Privacy & Fairness (45 min) | ||
| 16:00 | Challenges and Design Considerations for Finding CUDA Bugs Through GPU-Native Fuzzing | (Columbia University); (University of Pennsylvania); (Columbia University); (Columbia University) |
| 16:15 | Accelerating Fully Homomorphic Encryption at Scale: Lessons from Storage-Centric System Design | (University of California San Diego) |
| 16:30 | Health-Aware AI Inference Scheduling with Long-Term Fairness Objectives | (Rochester Institute of Technology); (University of California, Riverside); (California Institute of Technology); (University of California, Riverside) |
| Panel Discussion (60 min) | ||
| 16:45 – 17:45 | See Panelists below. | |
| 17:45 – 18:00 | Closing Remarks & Networking | Welcome Reception begins at 18:00 |
All talks: 12 min presentation + 3 min Q&A unless otherwise noted.
Panel Discussion
16:45 – 17:45
Sarah Cen
Carnegie Mellon University
ECE / EPP
Sarah studies accountability in AI systems — auditing, supply-chain transparency, and the legal and evidentiary challenges that arise when machine learning systems cause harm. She completed her PhD at MIT and a postdoc at Stanford HAI (Percy Liang, Daniel Ho), and brings a rigorous framework for translating between policy requirements and systems research.
Babak Falsafi
EPFL
Founding Director, EcoCloud | President, SDEA
Babak's research spans post-Moore server architecture and datacenter efficiency. He founded the Swiss Datacenter Efficiency Association (SDEA) — the only quantitative datacenter efficiency standard recognized by the IEA — and serves as the panel's practitioner voice on bridging the gap between research and enforceable industry criteria.
Ian McDougall
University of Wisconsin–Madison
PhD Student
Ian's systems work examines access and equity in computing infrastructure: open-source RISC-V toolchains for educational access, GPU cloud pricing fairness (Agora), and privacy-preserving GPU performance profiling. He brings a ground-level perspective rooted in concrete systems contributions and a critical eye toward the assumptions baked into current infrastructure.
Call for Papers
HotEthics'26 invites submissions that identify, analyze, or address the ethical dimensions of systems and architecture design. We welcome work spanning the full system stack — from chips and hardware to datacenter infrastructure and deployed ML pipelines. Topics of interest include, but are not limited to:
- Environmental impact of system design choices
- Systems and architectures for underserved, rural, and developing communities
- Bias introduced by computer architecture and systems
- Low-cost, low-power, and accessible device design
- Privacy- and security-aware system design
- Bias in machine learning for systems
- Benchmarking practices and their ethical implications
- System reliability, maintainability, and long-term societal impact
- Research methodology, evaluation norms, and incentive structures
Submission Guidelines
We invite submissions in four categories: completed research papers, industry and practice papers, work-in-progress papers, and short position papers that provoke, challenge, or motivate new directions. Submissions are not required to include formal quantitative or experimental results.
Short papers: up to 2 pages (excluding references).
Long papers: up to 4 pages (excluding references).
Formatting: double-column, 11pt font, US letter (8.5" × 11").
Review model: single-blind; submissions need not be anonymized.
Reviews are for acceptance decisions only — authors should not expect written feedback.
HotEthics'26 is non-archival. Authors retain full freedom to submit extended versions to conferences or journals. Accepted papers and slides will be posted on the workshop website.
Important Dates
- Submission deadline: January 30, 2026, Friday (AoE)
- Notification of decisions: February 13, 2026, Friday (AoE)
- Submission portal: HotCRP
- Workshop: March 22, 2026, Sunday (co-located with ASPLOS 2026)
