How AI Can Stop Unconscious Bias In Recruiting

A major feature of AI for recruiting is its ability to stop unconscious bias.

A recent lawsuit reminds us why we need to avoid bias during hiring: Palantir, a software startup, is paying $1.7 million to settle a racial discrimination lawsuit with the Department of Labor’s Office of Federal Contract Compliance Programs (OFCCP).

Similar to EEOC guidelines, as a federal government contractor, Palantir cannot discriminate based on race, color, religion, sex, sexual orientation, gender identity, national origin, disability or against military veterans.

unconscious bias and compliance

The Department of Labor accused Palantir of disproportionately eliminating qualified Asian applicants for engineering positions. The lawsuit alleges Asian applicants were routinely eliminated in the resume screening and telephone interview phases despite being as qualified as white applicants.

As just one high profile hiring discrimination case among many, this latest lawsuit drives home why compliance matters.

A lot of the blame for discrimination during hiring is placed on unconscious bias. Unconscious biases are “automatic, mental shortcuts used to process information and make decisions quickly” to which “everyone is susceptible.”

unconscious bias definition

In other words, unconscious bias is an ingrained human trait. That’s why some experts believe stopping unconscious bias requires an non-human solution: technology.

Here’s how AI for recruiting can help you avoid unconscious bias during hiring.

What unconscious bias looks like in recruiting

Research has found that resumes with English-sounding names receive requests for interviews 40% more often than identical resumes with Chinese, Indian, or Pakistani names.

The Palantir Asian discrimination case is a real life example of this type of bias.

Other common biases in recruiting include:

  • Similarity attraction effect: This is the tendency for people to seek out others who are just like them. “Opposites attract” turns out to be a myth: we tend to like individuals who are similar to us. Research on hiring practices has found that employers prefer candidates who are similar to themselves in terms of hobbies and life experiences, even though these similarities aren’t related to job performance.
  • Confirmation bias: This bias occurs when people favor information that confirms their beliefs and ignore or discount disconfirming information. Confirmation bias is one of the reasons why hiring managers ask different questions to different candidates during an interview. Because they tend to ask questions that confirm their unique beliefs about each candidate, this often results in comparing apples to oranges.
  • Halo effect: This bias occurs when we assume that because people are good at doing activity A, they will be good at doing activity B. In recruiting, the halo effect occurs when the hiring manager likes a candidate and uses that as a basis for assuming he or she will be good at the job rather than objectively assessing their skills and abilities.

Why unconscious bias is so hard to eliminate

The best-selling book Thinking, Fast and Slow explains the dual systems theory of the human mind. System 1 is fast, instinctive, and effortless. System 2 is slow, deliberate, and effortful.

Unconscious bias is a product of System 1 thinking. Because unconscious biases affect our thinking and decision making without our awareness, they can interfere with our true intentions.

Unconscious biases are so hard to overcome because they are automatic, act without our awareness, and there are so many of them: Wikipedia lists more than 180 decision making, social, and memory biases that affect us.

How recruiting AI reduces unconscious bias

AI for recruiting is the application of artificial intelligence such as machine learning, natural language processing, and sentiment analysis to the recruitment function.

AI can reduce unconscious bias in two ways.

1. AI makes sourcing and screening decisions based on data points

Recruiting AI sources and screens candidates by using large quantities of data. It combines these data points using algorithms to make predictions about who will be the best candidates. The human brain just can’t compete when processing information at this massive scale.

AI assesses these data points objectively – free from the assumptions, biases, and mental fatigue that humans are susceptible to.

A major advantage AI has over humans is its results can be tested and validated. An ideal candidate profile usually contains a list of skills, traits, and qualifications that people believe make up a successful employee. But often times, those qualifications are never tested to see if they correlate with on-the-job performance.

AI can create a profile based on the actual qualifications of successful employees, which provides hard data that either validates or disconfirms beliefs about what to look for in candidates.

2. AI can be programmed to ignore demographic information about candidates

Recruiting AI can be programmed to ignore demographic information about candidates such as gender, race, and age that have been shown to bias human decision making.

It can even be programmed to ignore details such as the names of schools attended and zip codes that can correlate with demographic-related information such as race and socioeconomic status.  

This is how AI software in the financial services industry is used. Banks are required to ensure that their algorithms are not producing outcomes based on data correlated with protected demographic variables such as race and gender.

AI still requires a human touch to stop unconscious bias

AI is trained to find patterns in previous behavior. That means that any human bias that may already be in your recruiting process – even if it’s unconscious – can be learned by AI.

Human oversight is still necessary to ensure the AI isn’t replicating existing biases or introducing new ones based on the data we give it.

Recruiting AI software can be tested for bias by using it to rank and grade candidates, and then assessing the demographic breakdown of those candidates.

The great thing is if AI does expose a bias in your recruiting, this gives you an opportunity to act on it. Aided by AI, we can use our human judgment and expertise to decide how to address any biases and improve our processes.

Ideal Blog Banner

Ji-A Min

Ji-A Min

Head Data Scientist at Ideal
Ji-A Min is the Head Data Scientist at Ideal. With a Master’s in Industrial-Organizational Psychology, Ji-A promotes best practices in data-based recruitment. She writes about research and trends in talent acquisition, recruitment tech, and people analytics.
Ji-A Min

Comments