Position of the German Social Insurance on the White Paper Artificial Intelligence
Dr. S-W – 06/2020
In
February the European Commission launched a debate on the safe and responsible
use of artificial intelligence. In their White
Paper it advocates a 'clear European regulatory framework', to protect
civil and fundamental rights, including data protection, privacy,
non-discrimination and consumer protection.
The umbrella organisations of the German social security system participated with a Statement at the consultation. They point out that there
is great potential in the development and use of artificial intelligence in a
European exchange of experience. However, as far as social security and the
services financed by it are concerned, supplementing already very general legal
principles with an equally general "adaptation of the existing EU legal
framework to take account of AI" does not bring additional legal certainty
or added value. However, this does not preclude the possibility of reviewing
certain individual provisions once more experience is available.
Especially
with regard to the use in the administrative area of social security, the
considerations about a potential use of AI are still in their infancy. For this
reason, there is no experience in this country of dealing with the imperative
of transparency, non-discrimination and final human decision making when using
AI. Often it will be shown that fundamental problems exist in advance and
independently of the use of AI, even independently of the use of algorithmic
systems. This becomes clear when looking at first experiences of public
administrations in dealing with AI and automatic decision-making systems in
other (not only European) countries. Therefore, a discussion on a responsible
handling of AI at the European level could provide very important indications
for the German discussion that is now getting underway.
According
to the position of the umbrella organisations of the German social security
system, the use of AI and corresponding European cooperation has great
potential for improving care, particularly in the health sector. In this
context, a comprehensive use of data is also necessary, including the use of
health data that are generated in the context of the provision of services by
the statutory health insurance scheme or that are financed by it. This includes
their use for training and learning of AI algorithms, as far as this serves e.g.
to improve the quality of care or for research purposes.
However,
it is important that we are cautious where data are concerned which not only
concern facts observable in the natural sciences, but also include a legal
derivation, e.g. in the question of whether an occupational disease or an
accident at work is present. There is no direct comparability of such data here
and it must first be examined whether a methodology for establishing
comparability can be developed.