Austria - profiling of the unemployed AMS/AMAS

What Happened

Austria’s Public Employment Service (AMS) developed and trialed an automated profiling algorithm named "AMAS" (Arbeitsmarktchancen-Assistenzsystem) to evaluate the employment prospects of jobseekers. The algorithm analyzed user data (such as employment history, age, education, and language skills) to assign risk scores that directly determined the level of career support and training resources allocated. The algorithm was heavily criticized after investigations revealed that it systematically deprioritized and delayed applications from older jobseekers and those from certain linguistic backgrounds, systematically under-allocating career support to vulnerable demographics based on flawed predictive indicators.

How MEG 1 would have acted

The profiling algorithm failed both the Dynamic Accuracy Index (DAI) and the Index of Safety and Responsibility (ISR) audits due to its systemic bias and violation of procedural fairness and employment access.

MEG 2 Analysis - Legal Framework

Agent level: N2 (management) - automated profiling and decision-making system used by a public authority to allocate public resources; directly impacts the right to work and social security.
Cause (6.1):
(a) system defect - algorithmic bias and flawed predictive indicators built into the allocation model (design defect), attributable to the developer.
(b) autonomous decision error - the system's decision to under-allocate essential funding based on flawed demographic proxies.
Attachment of liability: The public operator (AMS) is fully liable for the wrongful resource allocations caused by deploying an unverified, biased system.

Procedural Mechanism

6.2 (architectural human confirmation): Decisions involving the allocation of public funds or career support must require active, architectural human confirmation. The AI can only provide suggestions; a human officer must verify and sign off before final allocations are locked.
7.1 (independent forensics): The input parameters, the model's allocation weights, and the system's decision path must be securely logged in an independent ledger to allow for public audits.
9.4 (least privilege): The allocation system was granted direct authority over resource distribution without independent validation layers, violating safe operational parameters.