PublicAlgoBeta
All case studies

DK · 13 November 2024

Coded Injustice: Denmark's automated welfare fraud controls

Image: Amnesty International. All rights reserved by the owner.

Reported by
Amnesty International (Algorithmic Accountability Lab, with Amnesty Denmark)
Story language
EN · DA
Methods
  • FOI / records request
  • Interviews and surveys

Denmark's benefits agency runs about 60 fraud models, and it refused Amnesty's requests for a joint audit and full access to code and data. With redacted documents on four models and dozens of interviews, Amnesty still raised serious discrimination concerns: one model scores 'foreign affiliation', ties to countries outside the EEA, to prioritise investigations. It is a good example of building a credible case when code and data are refused. As far as we could find, nothing has changed since.

amnesty.orgCoded Injustice: Denmark's automated welfare fraud controlsAbout 60 AI and machine-learning fraud-control models used by Udbetaling Danmark (the public benefits payer) and run with the pension administrator ATP, with the contractor NNIT building the algorithms, to flag people for fraud investigation. Amnesty looked at four of them in part.Denmark · Reported by Amnesty International (Algorithmic Accountability Lab, with Amnesty Denmark) (opens in a new tab)

How they investigated

  1. FOI / records request

    Amnesty filed many requests with national and local employment and fraud agencies on technical design, governance and rationale. The agency supplied some redacted design documentation. It said it held no demographic data on risk classifications, and that high-risk case data is overwritten, so there is no history to test. It declined a collaborative audit and full access to code and data, citing sensitivity and the risk that disclosure helps fraudsters (GIJN).

  2. Interviews and surveys

    34 semi-structured interviews with officials, MPs, academics, journalists and affected people, two focus groups with people with disabilities, interviews with six refugee women and community leaders, and, with the women's shelter network LOKK, a survey of 25 shelter caseworkers on whether women in shelters could access benefits. Amnesty also attended a presentation by the agency's project team in January 2024. Finding affected interviewees was hard because of fear of reprisals.

Read alongside