An increasing number of companies are using artificial intelligence to screen resumes and streamline their hiring processes. In doing so, they may inadvertently be perpetuating gender bias, especially in traditionally male-dominated roles like sales and management.

Why is this happening? Because AI trains on historical data. If that data reflects biased hiring practices, then AI will “learn” based on these patterns and favor similarly biased profiles in its selection process.  

Say you are hiring for an in-house developer. Traditionally, most developers are men, so the tool will train on data sets that use “male coded” indicators (e.g. hobbies, phrasing, schools). The AI model associates these male-dominated patterns with greater likelihood of success. It may then ignore resumes from equally qualified women simply because the model is not trained to recognize “woman-coded” indicators.

What the Heck Is Male Coding?

Let’s use language as an example. What does “male-coded” vs. female-coded language look like? Let’s say two candidates submit resumes and cover letters for a software developer position at a large print company. Both are equally qualified, but their opening paragraphs might look very different.

A man might tend to write like this:

I am writing to apply for the software developer position at your printing company. With a proven ability to engineer scalable solutions and a strong analytical mindset, I’ve led multiple high-impact projects under tight deadlines. I bring a competitive edge, a results-driven approach, and technical leadership that drives performance and innovation. (AI generated)

A woman, on the other hand, is more likely to write like this:

I’m excited to apply for the software developer role at your printing company. I’m passionate about collaborating with cross-functional teams to create user-friendly, intuitive software solutions. My approach is thoughtful and detail-oriented, and I’m known for my patience, clear communication, and commitment to creating inclusive technology that meets real user needs. (AI generated)

Although both candidates may be equally qualified, an AI screening tool may ignore the second version because the tool is not trained on a woman-coded dataset and therefore may not associate keywords like “patience,” “cross-functional,” and “real user needs” with success in that position.

Even Names Result in Bias

AI-introduced gender bias can even result from someone’s name. A study from the University of Washington Information School found, for example, that resumes with male-associated names were prioritized 52% of the time, even for roles that are traditionally over-represented by women, such as HR managers (77% women) and secondary school teachers (57% women).

(Although not the topic of this article, it’s important to note that the inequities are even greater when it comes to race. The study found that resumes with White-associated names were selected 85% of the time, while resumes with Black-associated names were selected 9% of the time.)

Thanks to historical inequities, using AI screening tool can inadvertently perpetuate and even amplify the very gender biases many companies are working to address.

So what can you do?

• Choose design models tested for fairness.
• Track how each gender fares in your screening process. If you detect bias, go back and adjust the datasets or indicators used.
• Scrub job descriptions and resume scoring criteria of biased terms.

Don’t Overlook Job Descriptions

Remember that the same biases can exist (even inadvertently) in attracting applicants, too. For example, a traditional job description might open like this: “We're seeking a driven, strategic, and results-focused software developer to join our high-performance tech team at one of the region's leading printing companies.” This type of aggressive language can actually turn off many female job-seekers and dissuade them from applying.

Instead, adjust the language—and you can use AI to do this—to be more gender-neutral. Like this: “We’re looking for a thoughtful and skilled software developer to join our inclusive, forward-thinking team at a modern printing company.”

In other words, avoiding gender bias in hiring is a twofold process. It’s both in writing the job description and in the screening process. This can be the case, even if you’re not using AI. Many of these biases exist in our own brains, whether we realize it or not. The good news is, awareness is the first step in any type of correction.

Why should you care? First, because gender equity is the right thing to do. Second, it’s good for the bottom line. In fact, Fortune 500 firms with the best record of promoting women into high positions were 18% to 69% more profitable than the median firms in their industries.

Maybe it’s time to check the AI checker!