A Systematic Assessment of National Artificial Intelligence Policies: Perspectives from the Nordics and Beyond

By analysing 25 national AI policy documents, we identified differences in national prioritisation of AI related topics. For example, Sweden’s strategy has a strong focus on research, Japan on collaboration, and Serbia and France on ethical frameworks. Geographical clusters, such as the Northern Countries or the Anglosphere, are closely aligned in terms of ethical prioritisation, highlighting opportunities for future collaboration and geographical diversification of AI applications.

This entry summarises our NordiCHI 2020 paper “A Systematic Assessment of National Artificial Intelligence Policies: Perspectives from the Nordics and Beyond” by Niels van Berkel, Eleftherios Papachristos, Anastasia Giachanou, Simo Hosio, and Mikael B. Skov.

Figure 1. National governments increasingly develop their own artificial intelligence policies. Photo by Priscilla Du Preez.

The increasing impact of Artificial Intelligence (AI) on the everyday life of individuals and organisations has led many governments to take an increasingly active role in the assessment and support of new AI applications. Governments have stressed that AI applications should reflect national values, as for example seen in The White House’s push for “AI with American values”. In order to better understand the differences in how national governments position themselves on this emerging agenda, we set out to analyse national strategy and policy documents.

Through a systematic search, we identified 25 national policy documents. We applied a text analysis technique known as Latent Dirichlet Allocation to these documents to identify the topics that were discussed, identifying ten distinct topics;

1. Development strategy
2. Infrastructure
3. Private sector
4. Public sector
5. Data governance
6. Ethical framework
7. Education
8. Healthcare
9. Collaboration
10. Research.

We subsequently assessed the relative frequency with which these topics were discussed within each nation’s AI policy. Figure 2 highlights how the national policies are aligned to each topic, with closer proximity to a topic indicating that the topic was covered more extensively by the country. Countries at the extreme are quite unique in the focus they gave to specific topics. For example, Japan’s policy focuses more strongly on collaborations as compared to the other countries. Similarly, Spain focused more on development strategies, Serbia and France on ethical frameworks, Australia on the Private sector, and Sweden on Research than the rest of the countries.

Figure 2. Relation of a country’s national AI policy to the identified topics.

Strategy & Ethics

Whether or not AI behaves ethically and in line with our (moral) expectations is an often discussed topic both within popular media and academia. For a set of eleven ethical principles (e.g., transparency, sustainability), we assess the frequency with which they are discussed in each policy document. By grouping the documents in six geographical clusters, we aim to assess differences in the perspective of cultural or geographical clusters on the ethical concerns of AI. Figure 3 presents, for each of the six clusters, the average frequency with which the ethical principles are discussed. Countries within the Anglosphere cluster, for example, focus more heavily on ‘Justice & fairness’ than the entire collection of policy documents as a whole (called ‘corpus’). The South/East Asia cluster, on the other hand, highlights a stronger focus on ‘Sustainability’ and ‘Non-maleficence’.

Figure 3. Frequency with which ethical principles are discussed across six geographical clusters.

Reading the documents manually also uncovers some of the differences between countries. Ethic documents from the cluster of Northern European countries often refer to pre-existing cultural conditions as a base for expanding their AI efforts;

“Finland is known for its high level of citizen trust […] It is a privilege to step into the age of artificial intelligence from such an exceptional setting. At the same time, it practically obliges us to an active approach, understanding of the prerequisites of trust in the age of artificial intelligence […].”
— Final report of Finland’s Artificial Intelligence Programme

“Norwegian society is characterised by trust and respect for fundamental values such as human rights and privacy. This is something we perhaps take for granted in Norway, but leading the way in developing human-friendly and trustworthy artificial intelligence may prove a key advantage in today’s global competition.”
— Norway’s National Strategy for Artificial Intelligence

Our results highlight key differences between the analysed country’s (ethical) priorities regarding AI development. Using automated text analysis methods, we identify geographical clusters which align with existing cultural groupings. This highlights the opportunity for countries to cooperate on their AI research and development efforts with culturally aligned partners. Understanding the differences in perspectives towards AI technology between countries is a first step towards ensuring that AI applications align with the values of end-users cross-culturally.

“Europe and Denmark should not copy the US or China. Both countries are investing heavily in artificial intelligence, but with little regard for responsibility, ethical principles and privacy”.
— Denmark’s National Strategy for Artificial Intelligence

If you found this story interesting, please read the paper or contact one of the authors! We also recommend Jobin et al.’s paper “The global landscape of AI ethics guidelines”, which inspired our work.

Paper citation: N. van Berkel, E. Papachristos, A. Giachanou, S. Hosio, M. B. Skov, “A Systematic Assessment of National Artificial Intelligence Policies: Perspectives from the Nordics and Beyond”, Proceedings of the 11th Nordic Conference on Human-Computer Interaction (NordiCHI’20), 2020. doi: 10.1145/3419249.3420106.