Menu
Artificial intelligence (AI) has made its way into various fields, including the world of recruitment. Often, AI hiring tools assist with applicant screening, easing the burden of sorting through mountains of applications to find those with the most potential.
Overall, an estimated 55 percent of companies use AI during the hiring process. At times, their goal is to implement technology as a means of eliminating hiring bias. However, it doesn’t always go as planned.
While AI hiring tools aren’t often inherently biased, that can quickly change once data sources are introduced.
Company data on its top-performing employees isn’t inherently bias-free. If a company relies on that data for AI screening and the company’s best and brightest aren’t diverse, the AI won’t be inclusive.
One prime example of AI perpetuating bias involved a recruitment tool used by an e-commerce giant. As the AI screened technical job candidate resumes, it penalized applicants who used terminology indicating they were female. Along with the word “women,” the names of women’s colleges and similar items were viewed unfavorably by the AI.
The issue was that the AI was trained to identify patterns in resumes the e-commerce company received previously. Since most of the company’s technology candidates were male, the AI began favoring men.
Emotion recognition technology (ERT) – an AI-based approach to gauge an interviewee’s comfort level, confidence, and other emotional states – has also shown issues with bias. Emotion identification errors were far more likely when the technology was used during interviews with minorities, resulting in unfair hiring decisions.
For example, according to a study, some ERT technologies perceived black faces as angrier than white faces. As a result, reports generated by the ERT unfairly penalized black participants, which could lead to discriminatory hiring decisions.
While using AI tools is highly efficient, they may hinder corporate diversity since the source data that defines a perfect candidate likely isn’t bias-free. Similarly, flaws in ERT technologies could lead to biased assessments. In either case, relying too heavily on the technologies isn’t ideal.
When implementing these technologies, companies should take steps to reduce the level of potential bias. For instance, developing a dataset featuring a diverse talent pool could combat accidental bias within screening tools.
Additionally, keeping the human touch intact is essential. Even with concerns over unconscious bias among hiring managers, people can be made aware of their perceptions and make new decisions. AI will do only what its programming allows. If the programming is flawed, outcomes will be flawed.
Companies need to focus on creating fair hiring processes. Implementing blind hiring practices for screening is a straightforward starting point. Additionally, requiring additional training for hiring managers promotes awareness and gives them tools that they can use to make inclusive choices. In some cases, turning to a third-party recruitment partner could make a difference, particularly if the provider has a reputation for diversity and inclusion.
Ultimately, any step that reduces bias should be explored. That way, companies can develop a multi-faceted strategy that delivers meaningful results.
Feel free to share
