Comments on the EU White Paper on AI: A Regulatory Framework for High-Risk Healthcare AI Applications

4 Pages Posted: 25 Jun 2020

See all articles by Anastasiya Kiseleva

Anastasiya Kiseleva

Vrije Universiteit Brussel (VUB); CY Cergy Paris Université

Date Written: June 13, 2020

Abstract

The EU White Paper on AI mentions healthcare as one of the sectors where AI applications might pose high risks. These comments provide the vision on the regulatory framework for high-risk healthcare AI applications. The key takeaways concern transparency, preventing bias, safety and quality of AI applications used for medical purposes. I suggest that the transparency of AI shall not be equated to its explainability. To increase transparency and ensure the safety of AI healthcare applications, cooperation between manufacturers and users of AI (healthcare providers) shall be improved. I highlight that biased AI decisions alert on the inaccuracy of algorithms. While discrimination and stigmatization are difficult to identify and prove, especially for AI and especially for healthcare, controlling of AI’s accuracy is an efficient tool to prevent and mitigate biases in AI systems. Finally, I briefly compare the two regulations that might apply to AI tools in healthcare - the Medical Devices Regulation (EU) 2017/745 (MDR) and the In-vitro Diagnostic Medical Devices Regulation (EU) 2017/746 (IVDR). I conclude that the IVDR is more tailored to AI characteristics while it is more detailed and more focused on data quality and relevance. Thus, the IVDR rules can be a good starting point for clarifying and implementing a regulatory framework for high-risk AI healthcare applications.

Keywords: AI, EU White Paper on AI, High-Risk AI Applications, Transparency, Explainability, Bias, Discrimination, Stigmatization, MDR, IVDR, Medical Devices, Healthcare, Regulatory Framework on AI

Suggested Citation

Kiseleva, Anastasiya, Comments on the EU White Paper on AI: A Regulatory Framework for High-Risk Healthcare AI Applications (June 13, 2020). Available at SSRN: https://ssrn.com/abstract=3627741 or http://dx.doi.org/10.2139/ssrn.3627741

Anastasiya Kiseleva (Contact Author)

Vrije Universiteit Brussel (VUB) ( email )

Brussels
Belgium

CY Cergy Paris Université ( email )

paris
France

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
256
Abstract Views
1,192
Rank
217,401
PlumX Metrics