HotEthics 2024

Breakfast

Opening Remarks

Session 1: On a Mission to Reduce Emissions

•  

  The Sustainability Gap for Computing
  Lieven Eeckhout (Ghent University)

  Abstract: Computing is responsible for a significant and growing fraction of the world's global carbon footprint. Combating climate change and preserving sustainability in general is a grand challenge. This paper describes the sustainability gap for computing as a result of the socio-economic context (population and affluence growth) versus technology: the status quo in which we keep per-device carbon footprint constant would lead to a 5.4x gap relative to the Paris agreement within a decade. Meeting the Paris agreement for computing requires reducing the per-device carbon footprint by 15.5% per year under current population and affluence growth curves. Based on a select number of published carbon footprint reports, it appears that while (some) vendors indeed reduce the carbon footprint for (some) of their products, it does not seem to be enough to close the gap, urging our community to do more.

•  

  Silicon Efficiency in Post-Moore Servers
  Ali Ansari, Shanqing Lin, Ayan Chakraborty, Bugra Eryilmaz (EPFL); Mohammad Alian (University of Kansas); Babak Falsafi (EPFL); Michael Ferdman (Stony Brook University)

  Abstract: Server CPUs in the cloud have inherited their core microarchitecture from the desktop and mobile world, with performance primarily measured by single-core IPC. Furthermore, cores are integrated with large cache hierarchies within sockets and rely heavily on these caches to contain chip power envelopes, with little consideration given to utilization by workloads. Wasted silicon impacts both operational and embodied emissions in server platforms. In this work, we measure and compare silicon efficiency measured in performance per area and performance per watt of online and analytic services running on two x86 and an ARM server. We show that while x86 platforms offer higher single-core performance, the ARM server has the potential to achieve up to 2.5x higher socket-level performance per area and performance per watt than the x86 servers in the absence of system-level bottlenecks (e.g., memory or network bandwidth).

•  

  Intergenerational Embodied Carbon
  Aster Plotnik, Karthik Ganesan, Natalie Enright Jerger, Mark C. Jeffrey (University of Toronto)

  Abstract: The dawning of the age of accelerators has been a boon for computer architects and more broadly design and innovation in the computer hardware industry. However, as we spin and respin new accelerator designs, we must be cognizant of the broader implications of frequent redesign and accelerator churn. Specifically, we introduce the notion of lifetime carbon amortization and impact of legacy carbon due to frequent design refreshes. We argue for a more thoughtful and sustainable approach to accelerator design and deployment.

•  

  Towards Understanding the Carbon Impact in End-to-end Sensing Pipelines
  Harsh Desai, Sara McAllister, Nathan Beckmann, Brandon Lucia (Carnegie Mellon University)

  Abstract: The growth of sensing devices enables a wide range of previously untenable applications from sustainable agriculture to wildlife monitoring. At the same time, this growth necessitates considering the sustainability impact of these devices. Such devices capture sensor data, process it locally and radio-transmit it to the cloud via internet-enabled basestations. While some prior work has begun inspecting emissions at a device level, we need to understand the carbon impact of the entire sensing pipeline — from the sensing device to the cloud. Simply focusing on the device leaves several end-to-end impacts unexplored, both negative and positive. In this paper, we describe this end-to-end view of the sensing pipeline and show how one design decision in the sensing device affects the entire pipeline’s carbon impact.

Session 2: Thinking Outside the (Carbon) Box

•  

  The Environmental Impact of Forever Chemicals in Computing Systems
  Mariam Elgamal, Abdulrahman Mahmoud, Gu-Yeon Wei, David Brooks, Gage Hills (Harvard University)

  Abstract: The electronics and semiconductor industry are a prominent consumer of per- and poly-fluoroalkyl substances (PFAS), also known as forever chemicals. Computer designers and architects have an opportunity to reduce the use of PFAS in manufacturing semiconductors and electronics, including integrated circuits, batteries, displays, etc., which currently account for a staggering 10% of the total PFAS fluoropolymers usage in Europe alone. In this paper, we discuss the environmental impact of PFAS in computing systems, and how designers and architects can optimize for designs with lower PFAS-containing chemicals. We show that manufacturing a design with 16 nm technology node result in 15% less PFAS volume than manufacturing with a 28 nm lagging technology node due to area savings. We also show that manufacturing an IC at a 7 nm technology node using Extreme Ultraviolet (EUV) lithography uses 20% less volume of PFAS-containing chemicals, compared to manufacturing the same design at 7 nm technology node using Deep Ultraviolet (DUV) immersion lithography (instead of EUV).

•  

  Repair as Design: A Study on Engineering Student Attitudes and Experiences in Electronics Repair
  Esther Roorda, Emily Shilton, Sathish Gopalakrishnan (The University of British Columbia)

  Abstract: Addressing the rapid increase in global e-waste production requires a shift towards electronics repair, which not only mitigates e-waste but also reduces the carbon footprint associated with manufacturing new devices. While Right to Repair legislation aims to push manufacturers to design for repairability, 'repairability' is multifaceted, encompassing technical, practical, and socio-behavioral aspects. This study examines the attitudes and experiences of engineering students towards electronics repair, revealing gaps in both practical skills and awareness of environmental implications of their work, and a general lack of `repairabiliy' in existing consumer electronics. Despite their technical background, most students feel unprepared for repair and lack understanding of its environmental significance. The findings underscore the need for educational reforms that integrate repair skills and environmental literacy into engineering curricula to promote repairability in the design of personal electronics.

•  

  Electronic Waste Footprint of Computer Systems
  Pranjali Jain, Claire Pemberton, Ivan Hernandez, Mariana Rosillo, Samantha West, Jonathan Balkind, Timothy Sherwood (UC Santa Barbara)

  Abstract: Modern computer systems have an unprecedented environmental impact in the form of electronic waste at the end-of-life. Computer systems contain several hazardous materials whose improper disposal can lead to detrimental ecological and public health impacts due to their embodied toxicity. In this paper, we focus on developing methodologies to quantify the component-wise makeup of computer systems and assess their toxicity impact to inform more sustainable design choices.

Session 3: This Session is a No BrAIner

•  

  Towards Socially and Environmentally Responsible AI
  Pengfei Li, Yejia Liu, Jianyi Yang, Shaolei Ren (UC Riverside)

  Abstract: The sharply increasing sizes of artificial intelligence (AI) models come with significant energy consumption and environmental footprints, which can disproportionately impact certain (often marginalized) regions and hence create environmental inequity concerns. Moreover, concerns with social inequity have also emerged, as AI computing resources may not be equitably distributed across the globe and users from certain disadvantaged regions with severe resource constraints can consistently experience inferior model performance. Importantly, the inequity concerns that encompass both social and environmental dimensions still remain unexplored and have increasingly hindered responsible AI. In this paper, we leverage the spatial flexibility of AI inference workloads and propose equitable geographical load balancing (GLB) to fairly balance AI’s regional social and environmental costs. Concretely, to penalize the disproportionately high social and environmental costs for equity, we introduce Lq norms as novel regularization terms into the optimization objective for GLB decisions. Our empirical results based on real-world AI inference traces demonstrate that while the existing GLB algorithms result in disproportionately large social and environmental costs in certain regions, our proposed equitable GLB can fairly balance AI’s negative social and environmental costs across all the regions.

•  

  Is TinyML Sustainable? Assessing the Environmental Impacts of Machine Learning on Microcontrollers
  Shvetank Prakash, Matthew Stewart, Colby Banbury, Mark Mazumder (Harvard University); Pete Warden (Useful Sensors, Stanford University); Brian Plancher (Barnard College, Columbia University); Vijay Janapa Reddi (Harvard University)

  Abstract: The sustained growth of carbon emissions and global waste elicits significant sustainability concerns for our environment's future. The growing Internet of Things (IoT) has the potential to exacerbate this issue. However, an emerging area known as Tiny Machine Learning (TinyML) has the opportunity to help address these environmental challenges through sustainable computing practices. TinyML, the deployment of machine learning (ML) algorithms onto low-cost, low-power microcontroller systems, enables on-device sensor analytics that unlocks numerous always-on ML applications. This article discusses both the potential of these TinyML applications to address critical sustainability challenges, as well as the environmental footprint of this emerging technology. Through a complete life cycle analysis (LCA), we find that TinyML systems present opportunities to offset their carbon emissions by enabling applications that reduce the emissions of other sectors. Nevertheless, when globally scaled, the carbon footprint of TinyML systems is not negligible, necessitating that designers factor in environmental impact when formulating new devices. Finally, we outline research directions to enable further sustainable contributions of TinyML.

•  

  Towards Forever Access for Implanted Brain-Computer Interfaces
  Muhammed Ugur, Raghavendra Pradyumna Pothukuchi, Abhishek Bhattacharjee (Yale University)

  Abstract: Designs for implanted brain-computer interfaces (BCIs) have increased significantly in recent years. Each device promises better clinical outcomes and quality-of-life improvements, yet due to severe and inflexible safety constraints, progress requires tight co-design from materials to circuits and all the way up the stack to applications and algorithms. This trend has become more aggressive over time, forcing clinicians and patients to rely on vendor-specific hardware and software for deployment, maintenance, upgrades, and replacement. This over-reliance is ethically problematic, especially if companies go out-of-business or business objectives diverge from clinical promises. Device heterogeneity additionally burdens clinicians and healthcare facilities, adding complexity and costs for in-clinic visits, monitoring, and continuous access.

  Reliability, interoperability, portability, and future-proofed design is needed, but this unfortunately comes at a cost. These system features sap resources that would have otherwise been allocated to reduce power/energy and improve performance. Navigating this trade-off in a systematic way is critical to providing patients with forever access to their implants and reducing burdens placed on healthcare providers and caretakers. We study the integration of on-device storage to highlight the sensitivity of this trade-off and establish other points of interest within BCI design that require careful investigation. In the process, we revisit relevant problems in computer architecture and medical devices from the current era of hardware specialization and modern neurotechnology.

•  

  The Interplay of Computing, Ethics, and Policy in Brain-Computer Interface Design
  Muhammed Ugur, Raghavendra Pradyumna Pothukuchi, Abhishek Bhattacharjee (Yale University)

  Abstract: Brain-computer interfaces (BCIs) connect biological neurons in the brain with external systems like prosthetics and computers. They are increasingly incorporating processing capabilities to analyze and stimulate neural activity, and consequently, pose unique design challenges related to ethics, law and policy. For the first time, this paper articulates how ethical, legal, and policy considerations can shape BCI architecture design, and how decisions that architects make can constrain or expand the ethical, legal and policy frameworks that can be applied to them.

Session 4: Social(ly Conscious System Design) Commentary

•  

  Why Do We Need This Workshop?
  James McCauley (Mt. Holyoke College); Aurojit Panda (NYU); Scott Shenker (ICSI and UC Berkeley)

  Abstract: The purpose of this workshop is to bring together researchers who are interested in "designing computer systems and architectures in a socially responsible way'". This paper starts with simple question: Why do we -- as researchers who are interested in such issues -- feel the need for a special workshop devoted to this topic? To wit, why isn't social responsibility important in all technical conferences, and treated as just another metric along which systems are evaluated? Our response to this question consists of two obvious observations, that in turn lead us to the more fundamental question: How can this emerging "socially responsible" community have impact?

•  

  Can we optimize without specializing?
  Aurojit Panda (NYU)

  Abstract: This paper argues that a common approach to optimizing systems, one where we take advantage of specialized hardware, deployment and workload assumptions, comes at a high social and environmental cost. It discusses the cause of these costs, discusses why we do not consider them as a community, and suggests some approaches to address these costs.

•  

  Ethical Considerations of Benchmarking
  Victor Kariofillis, Jingyang Liu, Natalie Enright Jerger (University of Toronto)

  Abstract: Rapid advancement in the computer industry has sparked growing concerns about ethical aspects of computing. In this position paper, we explore an overlooked area: ethical dimensions of benchmarking practices in computer architecture. The selection of benchmarks embeds underlying ethical values into the final design. In light of this, we identify and discuss various shortcomings in current benchmark practices, point out their ethical implications, and make several proposals for how the computer architecture field can address them.

Session 5: Socially Conscious System Re-Design

•  

  The Once And Future Internet
  James McCauley (Mt. Holyoke College); Arvind Krishnamurthy (University of Washington); Tejas Narechania (UC Berkeley); Aurojit Panda (NYU); Scott Shenker (ICSI and UC Berkeley)

  Abstract: The Internet is the centerpiece of the world's communication infrastructure, and it touches almost all aspects of our lives. Thus, if there is any technology that should be designed in a socially conscious manner, it is the Internet. This short paper discusses what the Internet was, what it is now, and what it could become, all from the perspective of (quoting from this workshop's CfP) "potential ethical concerns arising from system and architecture design choices." Given the limitations on length, this paper focuses on identifying a few crucial, but often overlooked, social concerns related to the architectural design choices for the Internet, and leaves the details of our proposed solution to longer descriptions available elsewhere.

•  

  Towards Privacy-Preserving Audio Classification Systems
  Bhawana Chhaglani, Jeremy Gummeson, Prashant Shenoy (University of Massachusetts Amherst)

  Abstract: Audio signals can reveal intimate details about a person's life, including their conversations, health status, emotions, location, and personal preferences. Unauthorized access or misuse of this information can have profound personal and social implications. In an era increasingly populated by devices capable of audio recording, safeguarding user privacy is a critical obligation. This work studies the ethical and privacy concerns in current audio classification systems. We discuss the challenges and research directions in designing privacy-preserving audio sensing systems. We propose privacy-preserving audio features that can be used to classify wide range of audio classes, while being privacy preserving.

•  

  Demographic Bias in Data Center Scheduling Systems
  Sara Mahdizadeh Shahri (Carnegie Mellon University)

  Abstract: Modern web systems typically adopt a "performance-first" approach, where they prioritize sending quick responses to the end user to improve the Quality of Experience (QoE). However, putting performance first might make web systems introduce request priorities that may cause responses to be biased against certain user demographics. For example, to improve QoE, prior work often prioritizes requests that face lower network delays, implicitly causing biases against some requests that originate from rural areas.

  We make a case for how web systems must consider demographic bias as a key systems concern, to prevent discrimination against certain users. To this end, we systematically study and define demographic bias. We investigate whether modern web systems, especially scheduling systems, can (unintentionally) introduce demographic bias to improve performance, precipitating discrimination against certain user demographics. We detail a case study to show that demographic bias does occur in an open-source ML-driven scheduler that uses the Shortest Job First policy.

•  

  The Need for Equitable System Design: Lessons Learned From A Co-design Study with Low-Income Communities
  Stefany Cruz, Stephen Xia, Maia Jacobs (Northwestern University); Josiah Hester (Georgia Institute of Technology)

  Abstract: In this work, we call for careful consideration of the need for equitable systems design. Our work so far has involved co-designing wearables with members of low-income communities. From this experience, we have identified significant themes and lessons learned. We discuss how the computing systems community can play a key role in addressing low-income communities' challenges, such as lack of access to equitable health, safety, and environmental infrastructures through intelligent, ultra-low power, and battery-free system architectures.

Audience Panel/Discussion

As a group, discuss how the systems community can begin to consider ethical system design as a first-order consideration. (e.g., What are the challenges and opportunities in this space? How can we ensure that ethical considerations are not an afterthought, but rather a core part of the design process?)