Trelleborg welfare management automation
Trelleborg municipality in south Sweden was one of the first authorities in the country to automate the assessment and distribution of welfare payments.
Capitalising on a change in Swedish national law in 2018 that enabled the payment of welfare benefits without human intervention, Trelleborg municipality in south Sweden automated the assessment and distribution of welfare payments.
The system was a fully automated system using Robotic Process Automation (RPA) to assess and distribute payments. The local government claimed this would help them increase efficiency and reallocate resources elsewhere.
The approach taken, now known as the 'Trelleborg Model', illustrates how a seemingly well-meaning and apparently effective programme can go astray.
The impact was swift, with the municipality reporting that the number of people receiving social benefits had substantially decreased. And the number of case workers was reduced from 11 to 3 as individual assessments could now be made more or less instantaneously, having taken up to two days.
On the other hand, citizens complained of unfair assessments, little explanation of how the system worked, and limited ability to complain or appeal.
Researchers, journalists and other experts found themselves unable to access the system's data, code or model to assess its effectiveness or fairness, and freedom of information requests were summarily blocked on the grounds of trade secrecy.
Operator: Trelleborg Municipality
Developer: Trelleborg Municipality
Sector: Govt - welfare
Purpose: Optimise welfare payments
Technology: Robotic Process Automation (RPA)
Issue: Fairness; Employment - jobs; Privacy
Transparency: Governance; Black box; Complaints/appeals; Legal
Anne Kaun; September 2020: Suing the Algorithm
Anne Kaun, Lina Dencik; June 2020: Datafication and the Welfare State
Agneta Ranerup, Helle Zinner Henriksen; October 2019: Value positions viewed through the lens of automated decision-making: The case of social services
News, commentary, analysis
Published: February 2020
Last updated: December 2021