Can AI help eradicate unconscious bias?

Can AI help eradicate unconscious bias?

The call to greater diversity in the workplace has prompted a number of HR tech companies to develop solutions that detect unconscious bias – or prejudice deeply ingrained in the psyche that it is not always evident until closer analysis.

Artificial intelligence has become a key component in tools that help identify unconscious bias.

Tech companies such as, Joonko, and Textio are leveraging AI to detect these tendencies at various phases of talent acquisition and development.


Seattle-based app developer Textio combines human creativity and machine learning to produce ‘augmented writing’. The tool analyzes the gender bias in clients’ job descriptions, scoring their language on a scale of 100. The goal is to enhance the quality of writing so that job advertisements attract a diverse range of candidates.

“Gendered language isn’t always obvious or intuitive,” said Tim Halloran, brand experience director at Textio. But certain words such as ‘fearless’ and ‘enforcement’, which seem random, have been mathematically proven to skew the talent pool toward men.

Because of this, Textio has included a tone meter that indicates the likelihood of getting a response from one gender versus another. The spectrum of colors on the tone meter shows the writing’s ‘degrees of gender bias’. Dark blue suggests it is heavily biased toward men; dark purple toward women.

“If your tone meter is skewed toward the left or right, then look for the blue or purple highlights in your writing,” Halloran said. “Hover on those highlights with your mouse, and Textio will provide options for replacing that particular phrase.”

The tone meter works off the most current data on language use.

Talent intelligence platform claims women are 11% less likely to make it past an application review. The Mountain View startup addresses this hiring bias through blind screening in which the candidate pooling process removes any identifiers of gender from a profile.

Recruiters evaluate applications based solely on the individuals’ skillset, merits, specialization, and experience.

“We are fully compliant with EEOC and do not use age, sex, race, religion, disability, [etcetera] in assessing the fit of candidates to roles in enterprises,” Ashutosh Garg, CEO, told Jonathan Shieber of TechCrunch. The system can also generate gender-neutral job postings.

A lack of diversity in the talent pool may also be traced to a lack of diversity in recruitment datasets.

“Many of the biases people have in recruiting stem from the limited data people have seen,” Garg said.

To solve this, has mined publicly available information on the web along with internal data sources about the movement of talent within industries.

The platform sets out to build models of what an ideal workforce should look like and how to nudge companies into that direction. It relies on deep learning, the process by which a machine ‘learns’ from datasets and interactions, and predicts future outcomes based on those derived patterns. provides clients with a diverse talent pipeline that goes “beyond the few skills or companies they might know of, dramatically increasing their pool of qualified candidates,” said Garg.


Unconscious bias surfaces not only in business writing or computer datasets but even more so in everyday interactions in the workplace.

Joonko is an AI-powered app that identifies instances of bias in real time then provides immediate guidance on how to rectify them. For instance, an offhanded chat message about a person’s appearance might prompt the AI ‘coach’ to review the ‘potentially offensive words’. The user can then update the message before it is actually sent.

The app can be integrated into any existing recruitment, communication, and other professional management system. It can scan job descriptions, emails, and instant messages.

While AI can help detect instances of human bias, however, experts argue machine language isn’t necessarily devoid of its own loopholes. In place of unconscious prejudice is the possible emergence of algorithmic bias.

Related stories:
‘A robot took my job…’
AI experts command sky-high salaries
Revealed: top trends for recruiting in 2018