The EU aims to become a global leader in safe artificial intelligence. In its response to the European Commission consultation on the European approach for AI launched alongside a White Paper on Artificial Intelligence, focusing on actions to build an ecosystem of excellence and trust and on addressing questions of safety and liability, the European Public Health Alliance has called for a number of challenges to be tackled before comprehensive digitalisation of health and care is undertaken.
A multi-stakeholder dialogue needs to be created between digital and other sectors, including civil society, to improve awareness, education, and skills. End users need to be involved in the design, implementation and evaluation of AI solutions. Next to building up specific digital skills, it is important to focus on digital literacy, key in the health sector. Transparency and knowledge are crucial for ordinary people. Extensive coordination and involvement of end users is needed to improve the understanding of the benefits of using e.g. personal health data for the public good. A “Health in all Policies” approach also needs to be introduced in the technology sphere. Cross-sector bridges must be built to improve the understanding, meaning and deployment of AI solutions.
Given that patients, healthcare professionals and public health experts are the ultimate beneficiaries of AI and data-driven healthcare solutions (including pandemic surveillance technologies), public-private partnerships must involve civil society groups. The further development of AI in healthcare must not be driven by technology firms, to ensure it will bring benefits for everybody and address real needs. Research should address how AI can tackle health inequalities and improve access.
As a public health membership organisation, EPHA is especially concerned about any AI applications in healthcare and public health that could be operating on the basis of biased algorithms (not taking into account gender, ethnicity, etc.) and that do not take into account the broader determinants of health and histories of individual patients. We are also concerned that increased machine-generated decisions could potentially exacerbate existing health inequalities, discrimination and exclusion.
