AMAS — Austria's Banned-Then-Vindicated Jobseeker Profiling Algorithm
Austria
Austria's AMS scored every jobseeker into 3 employability tiers to ration training funds; researchers found a −0.14 penalty for being …
Poland · Warsaw · See the Poland profile
From 2014, ~350 Polish labour offices sorted every unemployed claimant into 1 of 3 algorithmic 'profiles'; research found 38% of offices had nothing to offer the neediest tier. Poland's Constitutional Tribunal struck the scheme down in 2018; it was abolished in 2019.
Starting in 2014, Poland's public employment service network — roughly 340–350 local Powiatowe Urzędy Pracy (district labour offices) nationwide — used a mandatory algorithmic system to sort every newly registered unemployed person into one of three 'profiles' based on 24 data points (8 collected at registration, the rest from a structured interview conducted by a caseworker at a computer terminal). Profile I ('active') jobseekers were expected to find work with minimal help; Profile II ('supported') received job-search assistance; Profile III ('distanced') were judged to have complex social or motivational barriers and were, in principle, eligible for more intensive interventions.
Independent research (Centre for Internet & Human Rights; corroborated by AlgorithmWatch's Automating Society reporting) found the three tiers split roughly 2% / 65% / 33% nationally, and that caseworkers overrode the algorithm's classification in only 0.58% of cases — meaning it was, in practice, nearly determinative. The same research found that 38% of labour offices were unable to offer any of the ten designated support types to their Profile III claimants — the group ostensibly most in need — because the specialised programmes were too costly or logistically difficult for smaller offices to run.
Poland's Human Rights Commissioner referred the scheme to the Constitutional Tribunal, arguing that the scope of personal data used to profile citizens should have been set by parliamentary statute rather than by ministerial regulation. The Tribunal agreed in a 2018 ruling, finding the delegation unconstitutional. The government subsequently passed an amendment abolishing the profiling system, which took effect in December 2019, ending five years of algorithmic triage of unemployment benefits nationwide.
Read the full analysis: https://cihr.eu/can-an-algorithm-hurt/
Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.
Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.
Where this practice's information was retrieved from, and when.
Austria
Austria's AMS scored every jobseeker into 3 employability tiers to ration training funds; researchers found a −0.14 penalty for being …
Ireland
Since 2012, Ireland's DSP has used a statistical model, PEX, to flag jobseekers at risk of long-term unemployment; a 2019 …
Belgium
Since ~2018, Flanders' employment service VDAB has used a Random Forest ML model, "Next-Step", to predict jobseekers' re-employment odds and …
Switzerland
SPSP integrates AI-powered genomic analysis to detect pathogen variants across Switzerland. Mandated by the Federal Office of Public Health since …
Open full copilot Grounded in cited practices — always check the sources.