With the regulation of artificial intelligence (AI) now a central issue with theAI Act, and in the light of European texts already in force such as the DSA (Digital Services Act) and the DMA (Digital Market Act), the contribution of the scientific community to this debate is proving necessary. But what are the regulator's needs and expectations, and how can scientific methods and approaches help to implement these complex regulations? What dialogue needs to be established, and what challenges will we have to face collectively to guarantee effective protection for citizens and respect for European values, while preserving the economic fabric and promoting innovation?
These were the questions addressed at the conference ‘Digital regulation: the contribution of science’, organised by Inria on 21 October at the Maison Irène et Frédéric Joliot Curie (Brussels).
"In recent years, the European Union has adopted a series of often pioneering regulations in the digital field. Such is the case of the DMA (Digital Market Act), which recently came into force, to prevent asymmetries and possible abusive and anti-competitive behaviour by ‘gatekeepers’, the major digital platforms capable of imposing their rules on their suppliers. Organising this seminar in Brussels gave us the opportunity to showcase the work of our scientists and French teams on a high-stakes subject that falls mainly within the remit of the European level‘, explains Marie-Hélène Pautrat, Director of European Partnerships at Inria, before adding ’The setting of the Maison Irène et Frédéric Joliot-Curie, a platform for French players in higher education, research and innovation, located in the very heart of the European quarter of Brussels, offers us a well-identified space to showcase the work of French teams and to exchange ideas with EU bodies and our European partners in Brussels’.
Building a dialogue between research and regulatory authorities
Through presentations by experts such as Nataliia Bielova, Stephan Lewandowsky, Jean Michel Loubes, Debabrota Basu, Erwan Le Merrer, Benoît Rottembourg, and Luca Aguzzoni (Head of Market Investigations, DG Connect, European Commission), This half-day of scientific presentations and discussions provided an opportunity to present the latest scientific advances and compare them with feedback from experience and the expectations of the European regulator with regard to the auditability of digital systems and artificial intelligence, understanding the risks associated with online platforms, thefairness of algorithms and the danger of ‘dark patterns’.
The final round table provided an opportunity to better define the potential forms of cooperation between scientists and the European regulator. In particular, it identified the forms of collaboration between scientists and regulators that work effectively and the obstacles that sometimes persist between these worlds, which have very different timeframes. Among the obstacles identified, the difficulty of bringing together multidisciplinary profiles - such as lawyers, economists, cognitive scientists, data scientists, computer scientists and web designers - within the same laboratory is a major challenge. The regulator therefore has to deal with the need to mobilise a wide range of expertise. Another difficulty is the mapping of skills and vocabulary between the terminologies used by the regulator and the research themes explored by the scientists. This alignment is not always easy.
One successful approach has been to ‘immerse’ scientists, for example by seconding them to an authority for a long period of time, which helps to capture this specific temporality and to bring the issues together while comparing points of view.
Lastly, access to data remains a key issue, not least because of the constraints associated with business secrecy and the protection or opacity of platforms. Recent legislation should, however, encourage better access to this data, which is a critical resource for researchers.
Verbatim
The feeling that emerges from this dialogue under construction between the world of research and that of the authorities in charge of regulation is that it is more necessary than ever in a context where recent ‘innovations’ in Artificial Intelligence have essentially come from the private sector and therefore need to be better understood by the rest of society. As one of our speakers paraphrased it, ‘you can only regulate what you can measure’, to which was added ‘you can only regulate what you can understand’. Given the conceivable systemic risks, there is an urgent need to measure, understand and regulate’.
Regalia Project Manager