THE NETHERLANDS. Several Dutch local authorities are using algorithms to predict if welfare benefit claimants are likely to be committing fraud, the NRC reported on Monday.
The algorithms are based on personal details which have been collected by government departments. Lekstroom in Utrecht province, a few parts of Zeeland and a council in Zuid-Holland province, which wants to remain secret, are all using the software already while another council in the east of the country plans to begin shortly.
Every year local authorities lose millions of euros to social security fraud – for example through people claiming to be single, which entitles them to more cash. And of the 100 potential fraudsters identified by the software, half were found to be cheating the system, the paper said.
The algorithm is ‘trained’ to recognise fraud by processing known cases based on date of birth, family situation, previous benefit history and information from the tax office, land registry office and car registration department RDW.
To remove bias – such as a focus on people living in certain streets or with a certain background – the results are then checked against a second system which looks at unusual behaviour.
The Dutch privacy watchdog Autoriteit Persoonsgevens told the NRC it does not yet know if the system conflicts with privacy legislation but points out that people whose information is being processed should be told in advance.
Next month, new European privacy legislation will come into effect which will stop people being profiled by computer, if this could have a significant impact on their lives.
However, because the final ruling on whether or not to stop benefits is made by a human being, the use of algorithms is unlikely to be affected, the NRC said.
This article appeared first on dutchnews.nl
DutchNews.nl is a provider of quality Dutch news and current affairs in English for an international audience. Some 25,000 people read DutchNews.nl every day, either online, through our daily digital newsletter or via our social media channels.