Sci Eng Ethics 27, 15 (2021).
In the development of governmental policy for artificial intelligence (AI) that is informed by ethics, one avenue currently pursued is that of drawing on “AI Ethics Principles”. However, these AI Ethics Principles often fail to be actioned in governmental policy. This paper proposes a novel framework for the development of ‘Actionable Principles for AI’. The approach acknowledges the relevance of AI Ethics Principles and homes in on methodological elements to increase their practical implementability in policy processes. As a case study, elements are extracted from the development process of the Ethics Guidelines for Trustworthy AI of the European Commission’s “High Level Expert Group on AI”. Subsequently, these elements are expanded on and evaluated in light of their ability to contribute to a prototype framework for the development of 'Actionable Principles for AI'. The paper proposes the following three propositions for the formation of such a prototype framework: (1) preliminary landscape assessments; (2) multi-stakeholder participation and cross-sectoral feedback; and, (3) mechanisms to support implementation and operationalizability.
In many areas, including AI, it has proven challenging to bridge ethics and governmental policy-making (Müller 2020, 1.3). To be clear, many AI Ethics Principles, such as those developed by industry actors or researchers for self-governance purposes, are not aimed at directly informing governmental policy-making, and therefore the challenge of bridging this gulf may not apply. Nonetheless, a significant subset of AI Ethics Principles are addressed to governmental actors, from the 2019 OECD Principles on AI (OECD 2019) to the US Defence Innovation Board’s AI Principles adopted by the Department of Defence (DIB 2019). Without focussing on any single effort in particular, the aggregate success of many AI Ethics Principles remains limited (Rességuier and Rodriques 2020). Clear shifts in governmental policy which can be directly traced back to preceding and corresponding sets of AI Ethics Principles, remain few and far between. This could mean, for example, concrete textual references reflecting a specific section of the AI Ethics Principle, or the establishment of (both enabling or preventative) policy actions building on relevant recommendations. A charitable interpretation could be that as governmental policy-making takes time, and given that the vast majority of AI Ethics Principles were published within the last two years, it may simply be premature to gauge (or dismiss) their impact. However, another interpretation could be that the current versions of AI Ethics Principles have fallen short of their promise, and reached their limitation for impact in governmental policy-making (henceforth: policy).
It is worth noting that successful actionability in policy goes well beyond AI Ethics Principles acting as a reference point. Actionable Principles could shape policy by influencing funding decisions, taxation, public education measures or social security programs. Concretely, this could mean increased funding into societally relevant areas, education programs to raise public awareness and increase vigilance, or to rethink retirement structures with regard to increased automation. To be sure, actionability in policy does not preclude impact in other adjacent domains, such as influencing codes of conduct for practitioners, clarifying what demands workers and unions should pose, or shaping consumer behaviour. Moreover, during political shifts or in response to a crisis, Actionable Principles may often prove to be the only (even if suboptimal) available governance tool to quickly inform precautionary and remedial (legal and) policy measures.