Poland PSZ unemployment scoring algorithm

Released: May 2015 

Can you improve this page?
Share your insights with us

Syriusz was an IT system used by Poland's public employment service Publiczne Służby Zatrudnienia (PSZ) that was upgraded in 2014 to incorporate a profiling mechanism that aimed to provide more personalised support for unemployed workers, and reduce inefficiencies and deliver better value for money at 340 job centres.

Based on points assigned during in-person registration, a survey and interview, the system placed unemployed workers into three categories. Each category determined the type of assistance they are entitled to, such as a job placement, vocational training, or  apprenticeship. 

The system was seen to suffer from a number of limitations and exposure to known risks that resulted in it being seen as poor value, ineffective, intrusive, unfair, discriminatory, and, ultimately, unconstitutional.

In 2015, Polish civil rights NGO Panoptykon Foundation produced Profiling the Unemployed in Poland, a detailed report focused on the system's lack of transparency and discriminatory character.

Lack of transparency

Poland's Ministry of Family, Labor and Social Policy and PSZ provided little information to job seekers or to third-party experts about how the profiling system worked, and refused to disclose important details such as the survey and scoring mechanism.

Furthermore, users had no ability to appeal decisions being made by the system and/or its human handlers.

Eventually, the ministry was forced to disclose the list of survey questions and the scoring mechanism after a lengthy, disputed Freedom of Information request submitted by the Panoptykon Foundation that ended up in court.

In 2018, Poland's Constitutional Court ruled the system was unconstitutional and that it should be closed by the end of 2019.

Operator: Publiczne Służby Zatrudnienia (PSZ)
Developer: Ministry of Family, Labor and Social Policy

Country: Poland

Sector: Govt - employment

Purpose: Assess unemployed worker support needs

Technology: Prediction algorithm
Issue: Appropriateness/need; Bias/discrimination - age, gender, disability; Effectiveness/value; Fairness; Privacy

Transparency: Governance; Black box; Complaints/appeals; Legal; Privacy