In order to avoid bias in AI, human resource will need to build fair and transparent data systems for diversified and high-quality talent acquisition.
Fremont, CA: In all the frenzy about the benefits of AI in discovering the best talent, organizations are waking up to the reality that the lost link was humans.
AI might be all powerful when it comes to automating many of repetitive tasks in HR, but it cannot always determine beyond the bounds of its algorithms. For instance, AI has the tendency toward inadvertent discrimination, often caused due to the inherent bias of the programmer. These biases often make themselves apparent when the AI goes to work—especially when it comes to recruitment. Even if we get rid of someone’s gender or ethnicity in the algorithms, the chances are that the machine could make prejudiced choices by spontaneously factoring in the candidate’s university or the area they are living in. Hence, it is imperative to build a strong ethics framework to mitigate AI’s bias in recruitment.
The face-recognition program is another example of the dilemma. An MIT study of three commercial gender recognition programs found that they had error percentage of up to 34 percent for dark skinned females, a rate roughly 49 times that for white males.
Why does this happen? Trained on inadequate, racially biased data, AI-enabled face recognition programs can often be erroneous. Lack of diversity in the personnel has been an incessant issue in the technology sector. There are ways in which HR could tackle such issues. The foremost thing to do in rectifying this issue is recognizing the problem. The following step that should be taken in addressing this bias is including the people against whom bias is directed, at all levels of the company that develop and administer the systems. Although onboarding on minority groups, women, and disabled people for some tech companies seems to be improving, it is not transparent as to how many of these people are in the position to make decisions that influence the biases.
It is impractical to hope for zero percent probability of conflicting impact from AI-driven talent acquisition algorithms. However, by bringing together consciousness and resolution designed to address such problems, human resource can re-invent future hiring processes.
