“Proven judgement and problem solving skills” and “Must have the ability to multi-task and have strong time management skills — able to work under pressure” were two of the desired characteristics in a job posting recently posted to hire a Re/code video shooter and editor.
Seems pretty basic and reasonable, right? I’d certainly want my future coworkers to have good judgment and manage their time well.
But it turns out the word “proven” and phrase “under pressure” tend to result in more male candidates, according to an analysis of thousands of job postings and which respondents were invited to interview.
And directive terms like “must” can be intimidating to many job seekers, regardless of gender. If a synonym is used, the analysis shows it will attract more applicants.
That’s according to Textio, a new Web-based tool that’s essentially a spell check for gender bias. You paste in the text of a job posting, and Textio highlights problematic phrasing and makes suggestions to help you attract more — and more diverse — good candidates.
Some of the tweaks are head-scratchingly tiny. Swapping out “exceptional” for “extraordinary” is statistically proven to attract more female applicants. (But while you’re at it, you should also insert “validated” instead of “proven.”)
Another pointer: Having too many listed bullet points often turns women candidates away.
According to Textio’s gender meter, the Re/code video editor posting was overwhelmingly male in tone. Textio didn’t spit out a number, but the little gender meter was almost completely over on the blue side.
That’s not a happy thing to learn, even though I would absolutely be willing to vouch that gender discrimination is not part of my Re/code experience. And I know we were just looking for someone fantastic at making videos.
But at least according to Textio, our job posting was an easy thing to fix.
I set to work making Textio’s suggestions — adding an equal opportunity statement, changing the word “candidate” to “you.” A couple minutes later, and with literally no substantive changes, the unconscious bias written into the posting and detected by Textio had been scrubbed.
Textio, which was developed by veterans of Amazon and Microsoft, is currently available by invitation and already being used by companies like Expedia. The company recently raised $1.5 million in funding from Cowboy Ventures, Bloomberg Beta and Upside Partnership.
Though gender has become a more frequent discussion topic in Silicon Valley of late, conversations about problems like bias and discrimination are often very separate from those about products and technology. With Textio, that’s not the case.
But Textio wasn’t built with a gender agenda, said CEO Kieran Snyder. Rather, it was built by her team of experts in machine learning and linguistic analysis with the idea of bringing real-time analysis to the writing process.
Snyder first started by analyzing Kickstarter campaigns. By simply comparing the text of Kickstarter pitches and their outcomes, Textio could predict with 93 percent accuracy whether a new project would be funded on day one.
“There are very few opinions involved in what we do,” said Snyder.
“All this is statistical, we don’t judge,” added her co-founder Jensen Harris, previously a well-known Microsoft executive who helped design Microsoft Office and other products. “In tech, people want to hire more women, but in nursing, they want to hire more men.”
Still, at this point the data available to Textio is limited. I’m not even sure you could call it “big data,” since it’s based on a combination of information scraped from job sites and follow-ups about hires contributed by an early list of participating companies.
Textio currently costs $59 per user per month, so it’s not inexpensive, even with discounts for company subscriptions. And many of its lessons could be absorbed by reading a list of do’s and don’t’s. At this point, it’s too early to say whether job posting language tweaks create better hires.
But Textio’s charm is in its interface, said Snyder and Harris. Copy and paste your job posting into its Web form, and Textio will overlay its analysis within seconds. The site channels massive amounts of analysis into each particular turn of phrase that’s highlighted in blue for skewing male, grey for being repetitive, or red for including jargon.
Job postings are the start, said Snyder. Eventually, Textio might like to do bias detection in other written areas of the workplace, such as performance reviews, the subject of so much argument in the recent Ellen Pao trial.
The other interesting part, said diversity consultant and workplace bias expert Joelle Emerson — who is doing research on Textio — is that it learns. “Rather than using static (and potentially outdated) research on what attracts certain types of candidates, it gives feedback based on what’s happening today,” Emerson said.
This article originally appeared on Recode.net.