From hiring candidates to evaluating employees, automated decision-making systems promise to make the world of employee management easier and bias-free. Of course, the news has been awash for years with examples of biased algorithmic tools, and New York recently enacted a law intended to regulate AI hiring.
With some of the largest companies preparing to outsource the human in Human Resources, smaller companies should take stock of the possibilities and limitations of these tools before following suit.
Technology doesn’t solve the problem of bias
You may be familiar with experiments from the early aughts that revealed racialized labor market discrimination. Given two identical resumes, the resume with a “white-sounding” name received 50% times more callbacks compared to the resume with the “Black-sounding” name.
Tools like the applicant tracking system are widely used to facilitate the hiring process due to their perceived impartiality. They don’t “see” an applicant’s identity, only relevant keywords, experiences, and skills. However, it neglects the fact that proxies for race, gender, class, and other identity markers are present, even if never explicitly indicated.
Where are these proxies present, you might ask? In the very metrics we use for assessing fit and merit.
While using automated systems may mitigate bias from proxies such as academic pedigree, some tools can penalize resume gaps, or a nontraditional employment history. Such metrics can discriminate against talented candidates with disabilities, those who’ve been incarcerated, or those who’ve taken parental or family leave.
When candidates and employees are flattened into a series of variables to be ranked and scored, inequities are inherently encoded. Some of these inequities stem from the politics of job titles, work assignments, or the under-resourcing (and subsequent underperforming) of departments and teams. Algorithmic tools could entrench these inequities by statistically favoring those who benefit from office politics.
Rather than doing the work to unlearn systemic biases present in the workplace, by outsourcing to AI we further marginalize the very communities these tools aren’t supposed to see. In the case of employment decision-making, this has the potential to lead to civil rights law violation claims.
The limits of efficiency
Automated decision making not only promises bias-free decisions, it also promises efficiency in managing complexity. While hiring managers, job seekers and employees can all attest to experiences with drawn out HR processes, what happens when we value efficiency over efficacy? It forces us to focus on what’s measurable — but more to the point, it forces us to focus on what’s economically measurable.
This has the consequence of long-term social initiatives being sacrificed for short-term gain, a topic I covered in a previous Portland Inno piece.
In a world where every problem has a business solution, it’s tempting to see automated decision-making systems as the efficient and impartial arbitrator to replace the slow and biased human element in employee management. Instead, use automated tools as a way to support the humans in recruiting and HR.
By balancing efficiency with effectiveness, your company can invest in places with the greatest potential for impact and innovation. From improving the job description and eliminating the six-round interview process, to leading with transparency, there are many ways in which bias can be more effectively addressed. What you save can be used to invest in diverse talent pipelines from job boards like PDXWIT, Diversify Tech, or the HBCU Career Center.
