In cooperation with the Polish Social
Insurance Institution (ZUS), the International Social Security Association
(ISSA) has conducted a virtual "AI in Social Security webinar".
kicked off with high standards: It wanted to find out how Artificial Intelligence (AI) affects the way in
which social security institutions provide their services.
This mainly involves
automating internal processes and communicating with the insured persons (chat
bots). While the latter application field seems to be making progress, the
other applications are still rather rare.
Estonia reported on a number of projects where AI is to be introduced.
For example, these include submitting job offers, medical triage and better
management of the risk of circulatory arrest.
Finally, the "KrattAI"
logo conceals the vision of an interoperable network of AI applications for
voice-based communication between citizens and authorities.
In Canada, algorithms trained by
"machine learning" were used to identify pensioners, who are entitled
to a means-tested basic pension, in a very special case scenario.
The aim was
to filter out a few thousand (suspected) cases from approximately 10 million
calls to call centres. Approximately 2,000 "vulnerable" cases in this
sense were actually detected. The success of this approach should be the
inspiration for further similar applications.
More fundamental considerations were
presented from Belgium, namely from the field of unemployment insurance.
Studying the use of AI started four years ago.
The decisive challenge was to digitise
the data, which had previously only been available in "analog" form,
through the use of text recognition systems and it was then made accessible for
"machine learning". However, the results have not yet been
convincing. The aim is apparently the further automation
of processes without compromising on quality.
The conference is accessible here; however, the
documents are only available to ISSA members.