Large employers increasingly rely on artificial intelligence to sort through job applications and choose who should reach the interview stage. Some argue that this makes recruitment fairer. Although AI can reduce certain human weaknesses, I largely disagree with this view, because these systems can repeat existing bias and unfairly reject capable candidates.
Admittedly, AI screening has some advantages in terms of fairness. Human recruiters who must read hundreds of applications grow tired and may judge the last applications more harshly than the first. They may also be influenced, consciously or not, by a candidate's name, age or university. A well-designed system, by contrast, applies the same criteria to every application, regardless of the time of day or the reader's personal preferences. In theory, this consistency should give each applicant an equal chance.
In practice, however, AI is only as fair as the data it learns from. Many systems are trained on records of the people a company has hired in the past. If those decisions favoured a particular group, for instance men in engineering roles, the software may learn to rank similar candidates more highly and quietly reproduce the same discrimination on a larger scale. Moreover, automated screening tends to reward applications that match a standard pattern. A talented candidate who has taken a career break to care for a parent, or who gained skills through unusual experience, may be filtered out before any human reads their application.
For these reasons, I do not believe that AI screening is fairer in its current form. It can be a helpful tool for managing large numbers of applications, but companies should regularly check their systems for bias and ensure that a person reviews the decisions, particularly when candidates are rejected.