Seattle-based Textio is introducing new AI-powered tools this week that detect and help employers remove language from job applications that can make older people, women, or people of color feel excluded. With exclusionary language removed, and increasing accommodation language included, companies can hire job candidates faster by reaching the broadest array of people. For companies like Nvidia, for example, its use of Textio has led to a higher number of women job applicants.
Textio uses natural language processing to color-highlight and interrupt bias communication in workplace text like job applications, LinkedIn job candidate outreach, or emails on Outlook or Gmail. Accommodation language- and ageism-fighting AI will be made available through the Textio connected language platform and through integrations with job applicant tracking systems like Greenhouse and soon Workday.
“You’re telling those people that you’re open to them coming and learning, it’s okay if you don’t have every skill on day one, and that inherently diversifies the people who will consider applying for a role,” Snyder said. “We know that people of color and women are less likely to apply to jobs if they don’t think they have 100% of the qualifications then white guys are, so if you can use language that communicates that it’s not just about what you’ve done, but your appetite for learning. You inherently reach that broader set of people who might become more confident to apply to that role.”
Snyder says reducing ageism and optimizing for inclusivity will be important in an economy turned upside down by the coronavirus pandemic. The United States Labor Department today reported that nearly 3.3 million people made unemployment insurance claims — a new record and an increase of 3 million in the past week. A poll earlier this month found 1 in 5 U.S. households lost work due to coronavirus. COVID-19 layoffs could fall particularly hard on African-Americans and Latinx Americans.
Among new Textio features is a graph for recognizing how certain words appeal to people by age. The graph demonstrates, for example, that snacks at work are more appealing to applicants in their 20s or over 60, and less appealing to people in their 30s. But mentioning benefits for dependents is more attractive for applicants in their 30s or 40s.
Snyder said Textio attempts to block platform users from targeting particular age or demographic groups by creating job applications optimized for maximum reach.
“The whole point of the implementation is that it takes you to age-inclusive, it takes you to neutral, so it helps you remove language that may be adversely impacting one age bracket or another,” she said. “In the era of COVID-19, as a lot of the impacted community are older folks, sort of age inclusivity around how we’re building roles that support community and have those lenses on their data is really important.”
Textio’s platform first started by removing language from applications that was tied closely to gender. The company recently upgraded that feature after working with nearly 500 enterprise customers. Race and ethnicity bias detection AI is also part of the Textio roadmap.
Building this sort of AI that seeks to reduce adverse impacts for any age group started with experiments, asking thousands of people from different age groups to share their feelings about sample language. Textio then built an initial data set and started working with companies collecting information about job candidates, and the company also sought feedback from advocacy and nonprofit groups.