Governing AI in the EU: public perception, policy dilemmas, and the path forward

The European Union (EU) has the stated goal of reaping “the full benefits of first-mover adopter advantages” for artificial intelligence (AI). In seeking to achieve this policy objective, democratic theories of policy implementation suggest that the likelihood of implementation success is improved when administrators and officials understand the unique demands and characteristics of those they govern. Therefore, we ask, what groups of EU residents may be the least trusting of AI, and what potential remedies exist to address AI trust deficits? Specifically, we test the association between an individual’s social class, views of government, engagement with science and technology, and their perceptions of the impacts of AI on society and the economy. We examine these relationships using data from Eurobarometer 95.2. Our findings indicate that individuals with positive perceptions of EU-level governments and those in higher social classes tend to have more favorable views of AI. Additionally, those who attend science museums and meetings about science and have a preference for non-expert involvement in science are associated with more positive views of AI. These findings highlight how both class divisions and differing perceptions of governments, who act as AI regulators, are associated with divergent views of AI. These trust deficits pose a potential problem for the implementation of existing EU AI policies. However, by transparently engaging citizens in meetings on the topic of AI while also including non-experts in the development of AI use cases, our findings suggest governments may be able to improve views of AI.