Bridging the Gap: Integrating Ethics and Environmental Sustainability in AI Research and Practice

Ethics and environmental concerns in AI are interconnected, not separate issues: The paper argues that carbon footprint analyses and ethical audits of AI models are typically conducted in isolation, missing critical insights that emerge when examining them together—such as how the pursuit of larger, more powerful models creates both accessibility barriers for researchers and environmental harms.

Scale in AI development creates compounding problems: The pursuit of increasingly large AI models contributes to two parallel harms: they become prohibitively expensive for most researchers to access or develop, and they generate substantial environmental costs that are often overlooked in ethical evaluations.

Current model audits are incomplete: Existing approaches to evaluate AI model performance and fairness largely fail to account for environmental impacts, meaning organizations may be making decisions about which models to deploy without understanding their full societal and ecological costs.

Solutions require integrated evaluation frameworks: The paper proposes that responsible AI practices should evaluate generalizability, transparency, evaluation methods, and equity alongside environmental sustainability metrics, rather than treating these as separate concerns.

Best practices should combine ethics and sustainability: The authors recommend that AI research and practice adopt integrated approaches that deliberately connect ethical considerations with environmental impact assessments from the outset, rather than addressing them as afterthoughts.