Highlights of the VEIL

In this Virtual Ethical Innovation Lecture, Larissa Bolte, research assistant at the Sustainable AI Lab, of the Institute of Science and Ethics at University of Bonn, gives an introduction into the concept of an Ethics of Desirability. She elucidates how such an innovative concept can help overcome methodological limitations, associated with guideline ethics and the (positivist) fixation on individual AI applications. The concept amplifies the ethical discussion and explicitly refers to wider social and environmental concerns. It is also connected to the broader debate on sustainability.

What is Ethics?

Starting by laying out the principles of ethics, which is assumed as an argumentative and sound discussion, Larissa Bolte emphasizes their dynamic and pluralistic character. Besides this openness, in history, ethics is the expression of reasoned reflections on what shall be good for humans, what is an adequate behavior, and, specially, how human interactions can be regulated for the best (for all or for the individuum). Such discussions are also present in the field of Artificial Intelligence (AI). (Though, the difference is that AI may be understood as a non-human actor, which is a new case for philosophy.)

What is AI Ethics?

Today, AI ethics is debated not only by the classic academic contributors in practical philosophy but additionally by public institutions and private tech-companies. Summing up, six principles are common in the discussions: Transparency, Inclusion, Accountability, Impartiality, Reliability, Security and Privacy. Ethics guideline documents are established by all the mentioned actors and constitute a substantial part of the discourse.

Critiques on the Actual Debates

It has been criticized that these ethics guidelines generally

a) show ethics as a hindrance to development,

b) still need further interpretation layers, and

c) do not identify the hidden costs.

Larissa Bolte develops this last point in more detail. Hidden costs, in an ethical sense, appear as indirect or side-effects. They stay hidden as long as the focus stays on an individual application. These unethical consequences are primarily not attributed to the discussed AI system but become visible when observed from a wider perspective. These costs can be of social nature, or appear in the ecosystem, e.g., as an extensive use of resources or the circulation of toxic materials required for the massive computing systems.

From an Ethics of Carefulness to an Ethics of Desirability

Following Thilo Hagendorff in the critique on just listing principles, Larissa Bolte argues, that the problem lies in the habit, that every AI application is discussed as an isolated artefact. Such an ethics of carefulness will always remain limited. To overcome this methodological obstacle, she proposes zooming out from the individual, maybe local implementation, and widening considerations towards the surrounding society, economy, and ecology. Environmental effects have to be regarded as well as current and future human needs. In this context, the notion of an ethics of desirability is introduced, taking into consideration the various aspects of sustainability. It is lanced against a limiting but widespread automatism within AI ethics, which just considers how to ethically deploy a given algorithm, instead of seeing the bigger picture.

Literature

Bolte, Larissa, Tijs Vandemeulebroucke, und Aimee Wynsberghe. 2022. „From an Ethics of Carefulness to an Ethics of Desirability: Going Beyond Current Ethics Approaches to Sustainable AI“. Sustainability 14 (April). https://doi.org/10.3390/su14084472.

Hagendorff, Thilo. 2020. „The Ethics of AI Ethics: An Evaluation of Guidelines“. Minds and Machines 30 (1): 99–120. https://doi.org/10.1007/s11023-020-09517-8.

Mittelstadt, Brent. 2019. „Principles Alone Cannot Guarantee Ethical AI“. Nature Machine Intelligence 1 (11): 501–7. https://doi.org/10.1038/s42256-019-0114-4.

Navas, Grettel, Giacomo D’Alisa, und Joan Martinez-Alier. 2022. „The role of working-class communities and the slow violence of toxic pollution in environmental health conflicts: A global perspective“. Global Environmental Change 73 (März): 102474. https://doi.org/10.1016/j.gloenvcha.2022.102474.

Strubell, Emma, Ananya Ganesh, und Andrew McCallum. 2019. „Energy and Policy Considerations for Deep Learning in NLP“. arXiv. https://doi.org/10.48550/arXiv.1906.02243.