TDWI | Training & Research | Business Intelligence, Analytics, Big Data, Data Warehousing

AI's Role in Pay Equity

How AI can help enterprises address gender-based pay discrepancies.

• By Adam Zoia
• November 27, 2018

Gender-based pay inequality is a measurable global phenomenon that is only beginning to be meaningfully addressed by law in the United States. Although aggressive pay equity laws have been recently passed in some U.S. states, not all employers are subject to these rules. Even for those employers under the specter of compliance risk, how to resolve internal issues of pay inequality is a decision that ultimately must be achieved by managers and HR personnel at each level of an organization.

For Further Reading: The Gender Pay Gap Remains Recruiting AI Talent Humans in the Loop for Machine Learning

Artificial intelligence can give an HR team powerful tools to adapt policies and evaluate employees, helping them achieve gender equality in the workplace.

Software tools, including those driven by AI, are often blind to the conscious and unconscious biases that can color a human's judgment. Efforts to eliminate the gender pay gap can -- and should -- begin at recruiting. AI tools can assist in writing job descriptions to attract a more diverse candidate pool (e.g., Textio Hire), connecting employers to candidates that would not typically be considered (e.g., Stella), and conducting job interviews on a neutral basis (e.g., HireVue).

Once an employee joins an organization, an AI tool can ensure they're being paid fairly relative to other comparable employees -- not just the ones with same title (e.g., CompIQ). [https://www.compensationiq.com/ Full disclosure -- the author is the CEO of CompIQ.] AI tools can also give insight into why an employee leaves to maximize the value of dollars spent on retention programs (e.g., CultureAmp).

However, software tools are not always bias-free. AI is subject to a particular type of bias known as uncoded bias. When AI platforms are fed with a data set that has been influenced by human bias, the AI algorithm inherits that bias. Consider a hypothetical recruiting AI platform that is fed two data sets: one with hired candidates (which happens to be overwhelmingly male) and another with rejected candidates. Even if the platform is designed to recommend candidates independent of gender, the system may end up more frequently recommending male candidates as these tend to have the same characteristics (e.g., schooling and experience) as those in the hired candidate data set.

This hypothetical demonstrates that AI tools require significant human intervention in order to return the desired outcomes. These systems must be thoughtfully engineered and vigilantly monitored, with an honest accounting of shortfalls and a continuous focus on goals. As with any other asset in an organization, AI tools must be well managed to achieve their intended purpose.

Committing to inclusive workplace practices is a worthy goal beyond being a compliance objective. A focus on cultivating and promoting a diverse workforce has been shown to have beneficial impacts on profits, and it also strongly communicates to employees and a company's clients that any person can be a great employee or a valued customer.

Perhaps in appreciation of the benefits of an inclusive workplace, a high-profile group of 16 employers has agreed to the Glassdoor Equal Pay Pledge: a pledge to pay their employees equally and make their salary data transparent and their findings public. Expect to see more employers taking these matters seriously and using AI tools to help them.