OECD moves forward on developing guidelines for artificial intelligence (AI)

OECD moves forward on developing guidelines for artificial intelligence (AI)

This article is part of the Forum Network series on Digitalisation

The Committee on Digital Economy Policy of the Organisation for Economic Co-operation and Development (OECD) develops public policy analysis and high-level recommendations to help governments and other stakeholders ensure that digital security and privacy protection foster the development of the digital economy.

In two weeks, the Committee will consider a proposed set of recommendations produced by a group of leading international experts while it discusses the first intergovernmental policy guidelines for artificial intelligence that will be submitted as a draft Recommendation to the OECD Council in May 2019.

This announcement follows the events from earlier this month, during which the AI Group of experts at the OECD (AIGO) completed its recommendations in meetings at the World Government Summit in Dubai.

World Government Summit in Dubai

I had the pleasure of chairing these exciting events full of experts who are passionate about artificial intelligence and ensuring that its benefits are widespread and inclusive. The enthusiasm was also shared by Andrew Wyckoff, the director of the OECD’s Science, Technology and Innovation Directorate, which has been spearheading the work on AI over the past few years. He said,

“The contributions of the AI expert group mark an important milestone in our efforts to ensure that governments and people share the economic and social benefits of AI and understand and minimise the risks.”

The recommendations of AIGO cover a broad range of public and private policy matters that are being challenged by artificial intelligence systems. They are based on a common understanding of AI concepts including: what is an AI system? What is the AI system lifecycle? Who are the stakeholders and the AI actors?

These recommendations identify five principles for responsible stewardship of trustworthy AI, namely: inclusive and sustainable growth and well-being, human-centred values and fairness, transparency and explainability, robustness and safety, and accountability.

They also include recommended national policy priorities for trustworthy AI, namely: investing in responsible AI research and development; fostering an enabling digital ecosystem for AI; providing an agile policy environment for AI; and building human capacity and preparing for job transformation and calls for international co-operation on policies for trustworthy AI.

From left to right: Karine Perset, Wonki Min and Anne Carblanc 

Throughout Summit, we heard from experts in the academic, public and private sectors alike, many of which are part of AIGO.

Osamu Sudoh, professor of economics at the University of Tokyo and an AIGO member, said Japan’s new strategy on AI will share many principles with the OECD. “The implementation of the OECD principles growing out of this discussion in the near future is very important,” he said.

Nozha Boujemaa, Chief Science and Innovation Officer at Median Technologies, former Director of research at the French Institute for Research in Computer Science and Automation and former vice chair of the European Commission AI High Level Group, said her participation in AIGO underscored the importance of collaboration between the EU and the OECD:

“We have a whole framework of shared values and shared objectives,” she said. “The convergence is important. And the OECD has a chance to have a wide impact on AI policy.”

The Institute of Electrical and Electronic Engineers (IEEE), a professional organisation representing 430,000 members worldwide, has worked on ethical design principles for practitioners. Konstantinos Karachalios, managing director of standards for IEEE and also an AIGO member, underlined that, in comparison to other fora, “The OECD brings the economic aspects and a commitment to global fairness that could lead to a new paradigm to guide the development and use of artificial intelligence systems”.  

From left: Jason Stanley, Ryan Budish, Jonathan Frankle (speaking), Adam Murray, Matt Chessen 

The AIGO’s recommended principles go beyond economics to cover issues like privacy, individual and worker rights, and safety and reliability in AI systems. The diversity of AIGO’s membership provided an opportunity to develop a broad consensus on critical issues.

For Pam Dixon, a group member and executive director of the World Privacy Forum, a public interest group, the public-private engagement provided an opportunity to articulate practices on data security and privacy that are needed urgently.

“I want to see these principles be adopted and matter,” said Dixon. “Artificial intelligence is not art on the wall. AI is the plumbing.”

AIGO member Christina Colclough, senior adviser to the UNI Global Union, a labour federation representing 22 million members, said the push for consensus was a critical aspect of the group’s work. “Technology knows no borders, so we all have to work together,” said Colclough.

The Committee on Digital Economy Policy (CDEP) established the expert group in May 2018 as part of a discussion on the need for the OECD Council to adopt policy principles that foster trust in, and adoption of, AI among Member countries. Nineteen countries participated in the group’s debate. They were joined by a cross section of representatives of the European Commission, UNESCO, labour and business associations, academia, civil society and professional organisations such as IEEE.

Developing AI principles is part of the OECD’s work over the past two years in examining the impact of new technologies on society through the multidisciplinary Going Digital and Next Production Revolution projects.

The OECD is also planning to launch in 2019 a policy observatory on AI, a participatory and interactive hub that would bring together the full resources of the organisation in one place, build a database of national AI strategies and identify promising AI applications for economic and social impact.

Related Topics

Artificial Intelligence

Please sign in

If you are a registered user on The OECD Forum Network, please sign in