Shaping the Ethical Future of HPC and AI

Elaine M. Raybourn

Social Scientist, Sandia National Laboratories

Ethical Failures in HPC/AI: A Wake-Up Call

“One way HPC differs from other fields is that we’re always thinking at scale,” Elaine says, highlighting the acute responsibility researchers bear. Their decisions — whether ethical, technical, or procedural — do not just affect individual projects; they can ripple across global systems, scientific outcomes, and society. “That’s the nature of the work — whether it’s data, computation, or impact — and our ethical considerations must match that scale.”

The stakes increase when research transitions from the lab to real-world application. A flawed simulation might seem like a technical issue on paper, but errors can cascade into real-world harm. For instance, an unvetted AI model used in emergency response might misallocate resources during a crisis, or a biased training dataset could skew scientific conclusions that influence national health or environmental policy.

At HPC scale, minor missteps can be magnified, and researchers must wrestle with the weight of that amplification. “You start to realize that what you’re building could impact lives across continents or generations,” Elaine notes. “It changes the way you think. It’s not just about whether the code runs, it’s about whether the outcome is fair, accurate, and just.”

Elaine warns that the sheer scale of HPC can become a psychological barrier for those involved. “Because HPC deals with science at such a massive scale, individuals may feel they lack the agency to influence ethical decision-making,” she explains, cautioning that developers might see downstream ethical considerations as someone else’s job. However, as HPC’s influence grows and AI is applied at scale, ethical engagement must include everyone, from individual researchers to team leaders, and from institutions to the broader scientific community.

“It is far better to have a voice in shaping innovation than to avoid accountability,” Elaine stresses. “Instead of seeing ethics as an obstacle, we should see it as an opportunity to shape the future of HPC and AI in a responsible, meaningful way.” This shift in perspective, she explains, is essential, ethics should not be seen as a constraint but as a fundamental part of how workflows are designed, leading to stronger, more trustworthy technologies.

Addressing Bias and Fairness at Scale

Ethical mindfulness is especially critical when addressing bias and fairness in large-scale computation. The vast reach of HPC systems means that existing biases can be amplified, or entirely new ones can emerge, if left unchecked. When asked how the HPC community can ensure fairness in AI-driven research, Elaine highlights key foundational practices. “One of the most fundamental steps is carefully examining training data to ensure it doesn’t lead to misinformed or uninformed decision-making,” she says.

But the challenge extends beyond data quality alone. “It’s not just about whether a model performs well,” she adds. “It’s about what data it was trained on, what scenarios it wasn’t built to handle, and who gets left out of the design.” In high-stakes scientific and policy environments, even small gaps in representation or flawed assumptions can cascade into far-reaching consequences, shaping healthcare decisions, misrepresenting environmental risk, or reinforcing inequities in resource distribution. For Elaine, fairness isn’t just a matter of technical adjustments; it requires a broader shift in perspective. “We have to design with awareness of the real-world conditions our models will face, and the real people they’ll affect.”

This framework, adapted from the mathematical modeling cycle (A.C.T. Aarts, 2010), shows iterative HPC problem-solving — from defining the problem to applying real-world-informed models.

Source: Raybourn & Muollo, “Guidelines for Practicing Responsible Innovation in HPC,” HCII 2023.

Operationalizing Ethics for Scale

If scale is where both the promise and peril of HPC/AI lie, then the question becomes: how do we build systems — and cultures — that keep ethics front and center as complexity grows? For Elaine, the answer lies in moving ethics from abstract discussions into concrete action. “It’s not enough to theorize about responsible computing,” she says. “We must make it visible in our processes, planning, and day-to-day decisions.”

At Sandia National Laboratories, where she’s often the lone social scientist among technical engineers, Elaine has refined a process for helping researchers and developers become stewards who shape ethical outcomes from the ground up. She emphasizes that while ethical responsibility starts at the individual level, it cannot end there. Together with colleagues, they have developed a framework for thinking about ethics that leverages three modes of advocacy.

“The first is self-advocacy — actively engaging in ethical discussions and training opportunities to develop a personal ethical mindset,” she explains. “Next is individual advocacy — serving as a role model within your team, promoting ethical guidelines, and encouraging open dialogue. Finally, there’s system advocacy, elevating the conversation to the institutional level, influencing policies, engaging with industry leaders, and advocating for ethical frameworks that shape the broader research and development landscape.”

Her approach involves using discussion prompts — specially designed cards developed through years of collaboration — to help teams engage with ethical considerations throughout a project’s lifecycle. Through this practice, teams can build a habit of ethical reflection, ensuring that ethics is not treated as an afterthought but as a core design function that evolves alongside the science.

The Role of SC25 in Advancing AI Ethics

The SC Conference has emerged as a vital forum for ethical discussions in HPC and AI. “Last year was the inaugural ethics workshop,” Elaine said, noting that it evolved from earlier Birds of a Feather (BoF) sessions organized by colleagues Jay Lofstead and Jakob Luettgau. Those sessions, which began in 2019, are steadily gaining momentum.

This progression from informal BoFs to a dedicated workshop reflects a rising sense of urgency around ethics in HPC and AI. With its ability to convene leaders from academia, government, and industry, SC25 fosters interdisciplinary dialogue across the entire ecosystem.

Jay Lofstead and Jakob Luettgau kicking off the Ethics in HPC BoF — sparking tough questions, practical insights, and community-driven dialogue. Don’t miss your chance to be part of the conversation shaping responsible AI at SC25.

For Elaine and her collaborators, a workshop or a BoF is more than a session, it’s a space for reflection, conversation, and community-building. “We’re not just talking about what we can build or design,” she explains. “We’re asking what kind of future we want to create, and how we get there together.” SC25 provides an opportunity to expand that conversation further. 

Join Us at SC25 in St. Louis

Collaboration and continuous learning are essential to advancing high performance computing — and to ensuring its ethical and responsible growth. SC25 offers an unparalleled opportunity to expand your knowledge, connect with experts, and engage with the global HPC community.

Attendees will experience cutting-edge research and real-world applications through technical presentations, papers, workshops, tutorials, posters, and Birds of a Feather (BoF) sessions — all designed to showcase breakthroughs in AI, HPC, and emerging technologies. Among these, discussions on ethics and responsible computing, like those led by Elaine Raybourn and her colleagues, highlight the critical role of ethical reflection in shaping the future of the field. SC25 provides a unique platform where leaders from industry, research institutions, and academia come together to share insights and innovations shaping the future.

Join us for a week of discovery and collaboration at SC25 in St. Louis, November 16–21, 2025.

Registration opens in July—don’t miss the chance to be part of the world’s premier supercomputing conference!