Gold & Purves' framework brings AI's environmental impact under ethics review
New framework urges ethics boards to judge AI research by planetary limits—not just human subjects.
Gold and Purves argue that computationally-intensive research (CIR), especially in AI, carries significant environmental costs that are currently overlooked by institutional ethics boards. Their paper, presented at LIMITS 2026, identifies a policy gap: most Research Ethics Committees (RECs) and Institutional Review Boards (IRBs) lack clear criteria to evaluate the planetary impact of CIR. The authors propose three concrete reforms: (1) scoping criteria to define which CIR projects require ethics review based on computational resource use, (2) evidential criteria for reviewers to assess environmental trade-offs, and (3) a structured reflection method for researchers to weigh the value of their work against planetary limits.
The framework aims to shift ethics review from a narrow focus on human subjects to a broader consideration of ecological sustainability. By embedding environmental accountability into the review process, the authors hope to encourage reflexive practice among researchers and push the field toward more responsible resource allocation. The paper serves as a practical guide for universities and funding agencies seeking to update ethics policies for the era of large-scale AI and high-performance computing.
- Proposes scoping criteria to identify CIR projects that require environmental ethics review.
- Provides evidential criteria for ethics boards to assess environmental impact of compute-heavy research.
- Introduces a self-reflection method for researchers to evaluate their work against planetary limits.
Why It Matters
Gives institutions a concrete tool to hold AI and HPC research accountable for its carbon footprint.