While recent debates about societal aspects of Artificial Intelligence (AI) have largely focused on ethics principles and guidelines, questions of governance, politics and policies of AI have received less scrutiny. To address this gap, over more than three years I have been leading AI governance research programme to examine questions such as – What are the roles of the government in shaping AI development and use? And what are the roles of society? How and why AI strategies are developed and launched? And which ideas, interests and values do they prioritise and which ones do they leave out?

These questions are addressed in our analysis of 49 AI policy documents adopted by national governments, international organizations, consultancies, think tanks, and civil society organizations. An interdisciplinary conceptual framework developed for this research approaches AI as an emerging technology characterized by fast growth, considerable impacts, uncertainty, hype, and positive as well as negative expectations. This research on AI governance has contributed to the Human Brain Project (HBP) work on societal and ethical aspects of AI and led to a number of outputs. Here I would like to highlight two recent scientific articles resulting from this research. 

Governance of AI: the roles of the government & publics

Our recent article ‘Framing governance for a contested emerging technology: insights from AI policy’ (Ulnicane et al 2020) is part of a forthcoming special issue on governance of AI and robotics in Policy & Society journal and examines how AI policy documents frame governance as a way to resolve public controversies about benefits and concerns related to AI. In the context of recent hype surrounding the development and use of AI, many countries and organizations have launched their AI strategies. In this process, they have learned from each other and also from international assemblies such as the World Economic Forum and the Organisation for Economic Cooperation and Development. This has led to some similarities across the documents. They depict the benefits of AI not only as traditional contribution of new technologies to economic growth but also focus on the potential of AI to help in tackling Grand Societal challenges and achieving the United Nations Sustainable Development Goals.

AI policy documents assign diverse roles to the government in shaping socially beneficial development and use of AI. These include promoting AI, mitigating risks, and encouraging involvement of stakeholders in AI development and use. Focusing on diverse roles assigned to the government helps to move beyond traditional debates in technology policy of big government vs. small government and to develop more nuanced understanding of the roles of the public sector in shaping new technologies. Furthermore, AI policy documents assign an important role to multi-stakeholder forums bringing together diverse publics. While importance of involving civil society, users, marginal and vulnerable groups have been discussed in AI policy documents, in practice it has been challenging to ensure balanced representation and avoid capture of these forums by well-resourced and organised vested interests.

‘Good Governance’ beyond ethics guidelines & regulation

Another article, ‘Good governance as a response of discontents? Déjà vu, or lessons for AI from other emerging technologies’ (Ulnicane et al 2021) is part of a recent special issue ‘Artificial Intelligence & its Discontents’ published in Interdisciplinary Science Reviews journal and further examines the hopes and fears associated with AI in policy documents, how they compare to other emerging technologies, as well as proposed solutions and their limitations. Typical solutions suggested for realizing benefits and mitigating risks of AI are ethics guidelines and regulation. While both are often mentioned together indicating that they are closely related, careful reading of the documents reveals some interesting differences in attitudes towards ethics guidelines and regulation. While there is a lot of enthusiasm about ethics guidelines, attitude towards regulation is more cautious with caveats attached. In debates about AI, this difference in attitudes is associated with the so-called ‘ethics washing’ when powerful actors focus on ethics guidelines as a way to avoid or delay binding regulation.

Drawing on studies of other emerging technologies, this article highlights that there are more opportunities for developing ‘good governance’ than ethics guidelines and regulation. These include cooperation and coordination between the state and diverse non-state actors, international research collaboration, and Responsible Innovation approach.

Framing Artificial Intelligence Futures

In addition to scientific publications, this research has also contributed to educational, outreach and other activities within and beyond the Human Brain Project. It provided input into the recently launched HBP Ethics and Society team’s Opinion ‘Trust and Transparency in Artificial Intelligence’. And the video lecture ‘Responsible Artificial Intelligence: Ethics, governance, and policy’, part of the HBP online course ‘Research, ethics and societal impact’ is an example of education and training activities stemming from this work.

This research has also led to invited talks and contributions such as a recent guest seminar ‘Framing Artificial Intelligence Futures’ and contribution to the webinar to launch the 2020 Government AI Readiness Index report for Europe. Further outputs include organization of conference panels, presentations at workshops and conferences, and creation of international networks. More scientific publications and activities based on this work are forthcoming, stay tuned!

By Inga Ulnicane

References:

Ulnicane, I., W. Knight, T. Leach, B. C. Stahl & W.-G. Wanjiku (2020) Framing governance for a contested emerging technology: insights from AI policy, Policy and Societyhttps://doi.org/10.1080/14494035.2020.1855800

Ulnicane, I., D. O. Eke, W. Knight, G. Ogoh & B. C. Stahl (2021) Good governance as a response to discontents? Déjà vu, or lessons for AI from other emerging technologies, Interdisciplinary Science Reviews, 46:1-2, 71-93 https://doi.org/10.1080/03080188.2020.1840220

Leave a Reply

Your email address will not be published. Required fields are marked *