Is 2022 the Year that AI Ethics Takes Sustainability Seriously?
The field of artificial intelligence (AI) ethics has rightly been concerned with how automated systems negatively affect people, especially those from vulnerable and minoritized communities. Shocking examples of biased algorithms, invasions of digital privacy, and technological surveillance serve as cases in point. These kinds of abuses have led to high-level policy discussions, dozens of corporate AI ethics guidelines, and substantial empirical and theoretical scholarship examining the real and potential harms that AI poses for humans. But in the process of trying to identify and curtail unethical uses of AI, corporations and scholars alike have virtually ignored the impacts of AI on the environment and the relationship between AI and the pursuit of sustainable development. Thanks to the efforts of a small but increasing group of researchers, the tide is beginning to turn. But much more needs to be done if we want to direct AI towards noble causes like protecting the planet, addressing climate change, and achieving the UN Sustainable Development Goals (SDGs).
AI Ethics and the Environment: A Case of Curious Neglect
Jobin and others report that only 13% of a large sample of AI ethics principles mention sustainability. A further analysis by Owe and Baum indicates that only 6% of these principles directly value the environment in its own right. In the AI ethics literature, a mere handful of manuscripts (some unpublished) examine the environmental implications of AI. Much of the recent work on the ethical dimensions of AI utilize human-centered analyses or advocate for a human rights-based framework for AI governance. Nature and sustainability are almost entirely absent from the equation.
However, some analysts have begun to explore the ways in which AI interfaces with the SDGs. For instance, notable works by Vinuesa et al. and Sætra probe the ways in which AI might contribute to the massive societal project that is the Global Goals. These are welcome interventions that the AI ethics community would be wise to build upon, especially since the current sustainable development agenda is scheduled to run through 2030.
By contrast, the fields of environmental politics and environmental sociology have paid comparatively more attention to these concerns. Books such as AI in the Wild, Artificial Intelligence and the Environmental Crisis, and Ecology, Artificial Intelligence, and Virtual Reality approach the promises and perils of AI with a critical eye, wary of the hidden environmental costs of rapid technological deployment. Such findings place front and center the environmental challenges that AI poses.
International Law on AI: A New Hope
Although international law on AI remains in its infancy, a couple recent developments suggest that the regulatory ecosystem surrounding this technology may be more forward-thinking than either organizations or the AI ethics literature. First, the European Commission’s High-Level Expert Group on Artificial Intelligence affords pride of place to sustainability in its report on Ethics Guidelines for Trustworthy AI. Specifically, chapter 2 includes a section on “societal and environmental well-being” that recognizes “other sentient beings and the environment” as stakeholders in the creation of AI systems and encourages “[s]ustainability and ecological responsibility.” The report also emphasizes that AI research should be oriented towards contributing to global initiatives like the SDGs.
Second, UNESCO’s Recommendation on the Ethics of Artificial Intelligence, hailed as the “first ever global agreement on the Ethics of AI,” devotes three articles to environmental issues. Article 17 encourages using AI to promote “[e]nvironmental and ecosystem flourishing…for humanity and other living beings,” while article 18 highlights the importance of limiting the carbon footprint and exploitation of natural resources associated with the life cycle of AI systems. Article 31 mentions the need to assess the “environmental impact of AI technologies” in the context of global objectives related to sustainability, such as the SDGs. Although the document does not go into great detail as to how the environmental impact should be assessed or how policymakers might balance the benefits of AI with its environmental consequences, this agreement opens the door for further discussion about how the international community might integrate sustainability imperatives into regulatory regimes related to AI.
These supranational and international instruments signal a growing awareness of the linkages between AI and the environment and their relevance to the SDGs. Further mainstreaming of environmental imperatives into AI regulatory schemes will be necessary in order to translate these aspirations into action. Soft law declarations should culminate in the promulgation of environmentally-sensitive AI policies at multiple levels of governance.
Towards a Greener AI
While many AI ethicists have raced forward to demand AI that advances human interests and safeguards human rights, these calls only serve to reify the very anthropocentrism that has resulted in the current climate crisis. A minor chorus of voices (i.e. Bossert and Hagendorff; Owe and Baum; Ziesche) has begun to sound the alarm about the need to widen the scope of AI ethics to include considerations for non-human animals. Expansions of this kind are crucial to the task of eliminating blind spots in the field that have placed humans alone at the center of the moral universe.
At the same time, we must accept the special responsibility that humans have in determining how to use technology in ways that do not threaten the living and non-living worlds or our collective futures. We ignore the roles played by power and profit at our own (and the planet’s) peril. As Dauvergne warns, “artificial intelligence is never going to produce a sustainability revolution within the contemporary global order” and it “has no capacity to overthrow the entrenched interests that are exploiting people and nature.” The only way to protect the Earth and all its inhabitants from the worst consequences of widespread AI adoption is to move beyond an exclusively human-centric mindset. A greener AI ethics must transition away from notions of justice that privilege humans above all else to a justice that is planetary in scope. Let 2022 be the year that this necessary movement begins in earnest.
This article originally appeared in the report Global Trends in AI 2022: Food for Thought from GAIEI Experts, published by the Global AI Ethics Institute.