Question

Over the last decade, automated screening tools have quietly reshaped recruitment. Promising speed and objectivity, these systems scan thousands of applications in minutes, ranking candidates against patterns drawn from past hiring data. Their defenders argue that machines, unlike humans, cannot be swayed by a candidate's accent, alma mater, or appearance. Yet a growing body of research suggests the opposite: algorithms trained on historical records often inherit the very biases they were meant to eliminate. A system trained on a decade of promotion records from a male-dominated industry will learn, quite rationally, to prefer male-coded resumes. The result is discrimination with a veneer of neutrality — decisions that are harder to challenge precisely because they appear mathematical. Regulators have begun to respond. The European Union's AI Act now classifies employment-screening systems as high-risk, demanding audits, documentation, and human oversight. Critics, however, warn that compliance alone may create a false sense of security, since audited systems can still encode subtle proxies for protected attributes — a candidate's postcode, for instance, may stand in for ethnicity. The lesson is not that automation has no place in hiring, but that objectivity is a property of design, not of technology itself. The author's attitude towards automated hiring tools can best be described as: