Systems that were meant to detect misuse of the benefits scheme, mistakenly labelled over 20,000 parents as fraudsters.
Amongst the upheaval, little attention was brought to the fact that the tax authority was making use of algorithms to guide its decision-making.
In a report by the Dutch Data Protection Authority, it became clear that a ‘self-learning’ algorithm was used to classify the benefit claims.
Whether it's pure AI or simple big spreadsheet chopping off tens of thousands that "match some filter" fraud detection is rarely that simple.When services impact real human beings and false positives have consequences that fall on the least resourced it's important to follow through with human eyeballs and good channels of appeal.
In Australia the parallel mistake was the Robodebt Scheme:
On 29 May 2020, the Morrison Government announced that it would scrap the debt recovery scheme, with 470,000 wrongly-issued debts to be repaid in full.
Initially, the total sum of the repayments was estimated to be A$721 million, however in November 2020 this figure expanded to A$1.2 billion after the Australian government settled a class-action lawsuit before it could go to trial.
https://en.wikipedia.org/wiki/Robodebt_scheme