AI bias describes systematic distortions in AI systems that can lead to unfair or discriminatory outcomes. The cause is often skewed or historically biased training data – for example when a model learns from past selection decisions in which certain groups are under-represented.
In recruiting the consequences are particularly sensitive. If a model was trained on data with women under-represented in leadership, it may score women lower for leadership roles. A well-known example is an Amazon model that downgraded CVs containing "women" and was therefore retired.
Mitigations are technical and organisational: balanced training data, fair evaluation metrics, human-in-the-loop reviews, regular bias audits, transparent documentation and clear complaint channels. In Europe the EU AI Act prescribes additional safeguards for high-risk applications like recruiting.
Lunigi uses AI where it helps – finding matching roles – while keeping human decisions with candidates themselves.