Imagine a digital filter that accidentally hides great candidates just because of where they live or their gender. This work builds a set of tools to spot these hidden biases and fix them. It's like a safety check for AI to make sure everyone gets a fair shot at a job. It also provides a guidebook for developers to build these systems correctly from the start.
Companies using AI for hiring risk unintentional discrimination, leading to legal penalties and loss of talent. There is a lack of clear technical tools to detect these biases while remaining compliant with strict GDPR data privacy rules.
A set of bias detection tools, an equality monitoring protocol, and a software development guide for creating fair AI. It also includes three industrial integration blueprints for monitoring and ranking.
If you are an online recruitment platform dealing with potential bias in candidate rankings — this project developed fairness-aware rankings and interventions that ensure a diverse talent pool. This helps avoid legal risks and improves hiring quality.
If you are an AI auditing firm dealing with the complexity of GDPR and anti-discrimination laws — this project developed an impact assessment and auditing method. This allows you to certify that a client's AI system is compliant with EU regulations.
If you are a large enterprise dealing with the lack of transparency in automated hiring tools — this project developed actionable interpretability tools. This allows managers to understand why the AI recommended a specific person and correct errors.
Based on available project data, all outputs will be released as open access publications, open source software, and open datasets, suggesting no direct licensing cost for the software.
Yes, the project includes three demonstrators specifically designed as integration blueprints for industrial environments, including partners like Adevinta and RAND.
The project explicitly states that all outputs will be released as open source software and open access publications.
It provides an equality monitoring protocol and legal analysis specifically designed to manage tensions between GDPR and EU Non-Discrimination Law.
Based on available project data, the project runs until 2026-01-31, and it provides integration blueprints to streamline the process in industrial settings.
The consortium is well-balanced for commercial translation, featuring 14 partners across 7 countries. With a 21% industry ratio including market leaders Adevinta and RAND, the research is grounded in real-world data. The mix of 6 universities and 4 other organizations (including NGOs and auditing pioneers like AlgorithmWatch) ensures that the technical tools are validated against legal and ethical standards.
Contact Universidad Pompeu Fabra for access to the open source software and guidelines.
Contact us to find the specific open-source auditing tools developed by FINDHR for your HR tech stack.
