Algorithmic Bias

FoundationsGovernance, Safety and Ethics

Also called: Bias (Algorithmic)

Algorithmic bias is the systematic production of prejudiced or unfair decisions by an artificial intelligence system, disadvantaging specific demographic groups, customers or operational segments. It originates in historical human bias in the training data, in samples that do not represent the population, or in the design of the algorithm itself. In automated hiring, credit and insurance processes it carries legal liability, regulatory sanction, brand erosion and ethical failure.

In practice

Bias is the primary liability driver examined in compliance audits of customer-facing automated systems. The test is not whether the system was built with good intent but whether outcomes differ across protected groups, and that has to be measured. A target that cannot produce the measurement has not run it.

Not sure where your organisation stands?

Take the free AI-readiness diagnostic.

Start the diagnostic